This is a live mirror of the Perl 5 development currently hosted at https://github.com/perl/perl5
[perl #70802] -i'*' refuses to work
[perl5.git] / sv.c
CommitLineData
a0d0e21e 1/* sv.c
79072805 2 *
1129b882 3 * Copyright (C) 1991, 1992, 1993, 1994, 1995, 1996, 1997, 1998, 1999, 2000,
83706693
RGS
4 * 2001, 2002, 2003, 2004, 2005, 2006, 2007, 2008, 2009 by Larry Wall
5 * and others
79072805
LW
6 *
7 * You may distribute under the terms of either the GNU General Public
8 * License or the Artistic License, as specified in the README file.
9 *
4ac71550
TC
10 */
11
12/*
13 * 'I wonder what the Entish is for "yes" and "no",' he thought.
14 * --Pippin
15 *
16 * [p.480 of _The Lord of the Rings_, III/iv: "Treebeard"]
17 */
18
19/*
645c22ef
DM
20 *
21 *
5e045b90
AMS
22 * This file contains the code that creates, manipulates and destroys
23 * scalar values (SVs). The other types (AV, HV, GV, etc.) reuse the
24 * structure of an SV, so their creation and destruction is handled
25 * here; higher-level functions are in av.c, hv.c, and so on. Opcode
26 * level functions (eg. substr, split, join) for each of the types are
27 * in the pp*.c files.
79072805
LW
28 */
29
30#include "EXTERN.h"
864dbfa3 31#define PERL_IN_SV_C
79072805 32#include "perl.h"
d2f185dc 33#include "regcomp.h"
79072805 34
51371543 35#define FCALL *f
2c5424a7 36
2f8ed50e
OS
37#ifdef __Lynx__
38/* Missing proto on LynxOS */
39 char *gconvert(double, int, int, char *);
40#endif
41
e23c8137 42#ifdef PERL_UTF8_CACHE_ASSERT
ab455f60 43/* if adding more checks watch out for the following tests:
e23c8137
JH
44 * t/op/index.t t/op/length.t t/op/pat.t t/op/substr.t
45 * lib/utf8.t lib/Unicode/Collate/t/index.t
46 * --jhi
47 */
6f207bd3 48# define ASSERT_UTF8_CACHE(cache) \
ab455f60
NC
49 STMT_START { if (cache) { assert((cache)[0] <= (cache)[1]); \
50 assert((cache)[2] <= (cache)[3]); \
51 assert((cache)[3] <= (cache)[1]);} \
52 } STMT_END
e23c8137 53#else
6f207bd3 54# define ASSERT_UTF8_CACHE(cache) NOOP
e23c8137
JH
55#endif
56
f8c7b90f 57#ifdef PERL_OLD_COPY_ON_WRITE
765f542d 58#define SV_COW_NEXT_SV(sv) INT2PTR(SV *,SvUVX(sv))
607fa7f2 59#define SV_COW_NEXT_SV_SET(current,next) SvUV_set(current, PTR2UV(next))
b5ccf5f2 60/* This is a pessimistic view. Scalar must be purely a read-write PV to copy-
765f542d 61 on-write. */
765f542d 62#endif
645c22ef
DM
63
64/* ============================================================================
65
66=head1 Allocation and deallocation of SVs.
67
d2a0f284
JC
68An SV (or AV, HV, etc.) is allocated in two parts: the head (struct
69sv, av, hv...) contains type and reference count information, and for
70many types, a pointer to the body (struct xrv, xpv, xpviv...), which
71contains fields specific to each type. Some types store all they need
72in the head, so don't have a body.
73
74In all but the most memory-paranoid configuations (ex: PURIFY), heads
75and bodies are allocated out of arenas, which by default are
76approximately 4K chunks of memory parcelled up into N heads or bodies.
93e68bfb
JC
77Sv-bodies are allocated by their sv-type, guaranteeing size
78consistency needed to allocate safely from arrays.
79
d2a0f284
JC
80For SV-heads, the first slot in each arena is reserved, and holds a
81link to the next arena, some flags, and a note of the number of slots.
82Snaked through each arena chain is a linked list of free items; when
83this becomes empty, an extra arena is allocated and divided up into N
84items which are threaded into the free list.
85
86SV-bodies are similar, but they use arena-sets by default, which
87separate the link and info from the arena itself, and reclaim the 1st
88slot in the arena. SV-bodies are further described later.
645c22ef
DM
89
90The following global variables are associated with arenas:
91
92 PL_sv_arenaroot pointer to list of SV arenas
93 PL_sv_root pointer to list of free SV structures
94
d2a0f284
JC
95 PL_body_arenas head of linked-list of body arenas
96 PL_body_roots[] array of pointers to list of free bodies of svtype
97 arrays are indexed by the svtype needed
93e68bfb 98
d2a0f284
JC
99A few special SV heads are not allocated from an arena, but are
100instead directly created in the interpreter structure, eg PL_sv_undef.
93e68bfb
JC
101The size of arenas can be changed from the default by setting
102PERL_ARENA_SIZE appropriately at compile time.
645c22ef
DM
103
104The SV arena serves the secondary purpose of allowing still-live SVs
105to be located and destroyed during final cleanup.
106
107At the lowest level, the macros new_SV() and del_SV() grab and free
108an SV head. (If debugging with -DD, del_SV() calls the function S_del_sv()
109to return the SV to the free list with error checking.) new_SV() calls
110more_sv() / sv_add_arena() to add an extra arena if the free list is empty.
111SVs in the free list have their SvTYPE field set to all ones.
112
ff276b08 113At the time of very final cleanup, sv_free_arenas() is called from
645c22ef 114perl_destruct() to physically free all the arenas allocated since the
6a93a7e5 115start of the interpreter.
645c22ef 116
645c22ef
DM
117The function visit() scans the SV arenas list, and calls a specified
118function for each SV it finds which is still live - ie which has an SvTYPE
119other than all 1's, and a non-zero SvREFCNT. visit() is used by the
120following functions (specified as [function that calls visit()] / [function
121called by visit() for each SV]):
122
123 sv_report_used() / do_report_used()
f2524eef 124 dump all remaining SVs (debugging aid)
645c22ef
DM
125
126 sv_clean_objs() / do_clean_objs(),do_clean_named_objs()
127 Attempt to free all objects pointed to by RVs,
128 and, unless DISABLE_DESTRUCTOR_KLUDGE is defined,
129 try to do the same for all objects indirectly
130 referenced by typeglobs too. Called once from
131 perl_destruct(), prior to calling sv_clean_all()
132 below.
133
134 sv_clean_all() / do_clean_all()
135 SvREFCNT_dec(sv) each remaining SV, possibly
136 triggering an sv_free(). It also sets the
137 SVf_BREAK flag on the SV to indicate that the
138 refcnt has been artificially lowered, and thus
139 stopping sv_free() from giving spurious warnings
140 about SVs which unexpectedly have a refcnt
141 of zero. called repeatedly from perl_destruct()
142 until there are no SVs left.
143
93e68bfb 144=head2 Arena allocator API Summary
645c22ef
DM
145
146Private API to rest of sv.c
147
148 new_SV(), del_SV(),
149
150 new_XIV(), del_XIV(),
151 new_XNV(), del_XNV(),
152 etc
153
154Public API:
155
8cf8f3d1 156 sv_report_used(), sv_clean_objs(), sv_clean_all(), sv_free_arenas()
645c22ef 157
645c22ef
DM
158=cut
159
3e8320cc 160 * ========================================================================= */
645c22ef 161
4561caa4
CS
162/*
163 * "A time to plant, and a time to uproot what was planted..."
164 */
165
77354fb4 166void
de37a194 167Perl_offer_nice_chunk(pTHX_ void *const chunk, const U32 chunk_size)
77354fb4 168{
97aff369 169 dVAR;
77354fb4
NC
170 void *new_chunk;
171 U32 new_chunk_size;
7918f24d
NC
172
173 PERL_ARGS_ASSERT_OFFER_NICE_CHUNK;
174
77354fb4
NC
175 new_chunk = (void *)(chunk);
176 new_chunk_size = (chunk_size);
177 if (new_chunk_size > PL_nice_chunk_size) {
178 Safefree(PL_nice_chunk);
179 PL_nice_chunk = (char *) new_chunk;
180 PL_nice_chunk_size = new_chunk_size;
181 } else {
182 Safefree(chunk);
183 }
77354fb4 184}
cac9b346 185
d7a2c63c
MHM
186#ifdef PERL_MEM_LOG
187# define MEM_LOG_NEW_SV(sv, file, line, func) \
188 Perl_mem_log_new_sv(sv, file, line, func)
189# define MEM_LOG_DEL_SV(sv, file, line, func) \
190 Perl_mem_log_del_sv(sv, file, line, func)
191#else
192# define MEM_LOG_NEW_SV(sv, file, line, func) NOOP
193# define MEM_LOG_DEL_SV(sv, file, line, func) NOOP
194#endif
195
fd0854ff 196#ifdef DEBUG_LEAKING_SCALARS
22162ca8 197# define FREE_SV_DEBUG_FILE(sv) Safefree((sv)->sv_debug_file)
d7a2c63c
MHM
198# define DEBUG_SV_SERIAL(sv) \
199 DEBUG_m(PerlIO_printf(Perl_debug_log, "0x%"UVxf": (%05ld) del_SV\n", \
200 PTR2UV(sv), (long)(sv)->sv_debug_serial))
fd0854ff
DM
201#else
202# define FREE_SV_DEBUG_FILE(sv)
d7a2c63c 203# define DEBUG_SV_SERIAL(sv) NOOP
fd0854ff
DM
204#endif
205
48614a46
NC
206#ifdef PERL_POISON
207# define SvARENA_CHAIN(sv) ((sv)->sv_u.svu_rv)
daba3364 208# define SvARENA_CHAIN_SET(sv,val) (sv)->sv_u.svu_rv = MUTABLE_SV((val))
48614a46
NC
209/* Whilst I'd love to do this, it seems that things like to check on
210 unreferenced scalars
7e337ee0 211# define POSION_SV_HEAD(sv) PoisonNew(sv, 1, struct STRUCT_SV)
48614a46 212*/
7e337ee0
JH
213# define POSION_SV_HEAD(sv) PoisonNew(&SvANY(sv), 1, void *), \
214 PoisonNew(&SvREFCNT(sv), 1, U32)
48614a46
NC
215#else
216# define SvARENA_CHAIN(sv) SvANY(sv)
3eef1deb 217# define SvARENA_CHAIN_SET(sv,val) SvANY(sv) = (void *)(val)
48614a46
NC
218# define POSION_SV_HEAD(sv)
219#endif
220
990198f0
DM
221/* Mark an SV head as unused, and add to free list.
222 *
223 * If SVf_BREAK is set, skip adding it to the free list, as this SV had
224 * its refcount artificially decremented during global destruction, so
225 * there may be dangling pointers to it. The last thing we want in that
226 * case is for it to be reused. */
227
053fc874
GS
228#define plant_SV(p) \
229 STMT_START { \
990198f0 230 const U32 old_flags = SvFLAGS(p); \
d7a2c63c
MHM
231 MEM_LOG_DEL_SV(p, __FILE__, __LINE__, FUNCTION__); \
232 DEBUG_SV_SERIAL(p); \
fd0854ff 233 FREE_SV_DEBUG_FILE(p); \
48614a46 234 POSION_SV_HEAD(p); \
053fc874 235 SvFLAGS(p) = SVTYPEMASK; \
990198f0 236 if (!(old_flags & SVf_BREAK)) { \
3eef1deb 237 SvARENA_CHAIN_SET(p, PL_sv_root); \
990198f0
DM
238 PL_sv_root = (p); \
239 } \
053fc874
GS
240 --PL_sv_count; \
241 } STMT_END
a0d0e21e 242
053fc874
GS
243#define uproot_SV(p) \
244 STMT_START { \
245 (p) = PL_sv_root; \
daba3364 246 PL_sv_root = MUTABLE_SV(SvARENA_CHAIN(p)); \
053fc874
GS
247 ++PL_sv_count; \
248 } STMT_END
249
645c22ef 250
cac9b346
NC
251/* make some more SVs by adding another arena */
252
cac9b346
NC
253STATIC SV*
254S_more_sv(pTHX)
255{
97aff369 256 dVAR;
cac9b346
NC
257 SV* sv;
258
259 if (PL_nice_chunk) {
260 sv_add_arena(PL_nice_chunk, PL_nice_chunk_size, 0);
bd61b366 261 PL_nice_chunk = NULL;
cac9b346
NC
262 PL_nice_chunk_size = 0;
263 }
264 else {
265 char *chunk; /* must use New here to match call to */
d2a0f284 266 Newx(chunk,PERL_ARENA_SIZE,char); /* Safefree() in sv_free_arenas() */
2e7ed132 267 sv_add_arena(chunk, PERL_ARENA_SIZE, 0);
cac9b346
NC
268 }
269 uproot_SV(sv);
270 return sv;
271}
272
645c22ef
DM
273/* new_SV(): return a new, empty SV head */
274
eba0f806
DM
275#ifdef DEBUG_LEAKING_SCALARS
276/* provide a real function for a debugger to play with */
277STATIC SV*
d7a2c63c 278S_new_SV(pTHX_ const char *file, int line, const char *func)
eba0f806
DM
279{
280 SV* sv;
281
eba0f806
DM
282 if (PL_sv_root)
283 uproot_SV(sv);
284 else
cac9b346 285 sv = S_more_sv(aTHX);
eba0f806
DM
286 SvANY(sv) = 0;
287 SvREFCNT(sv) = 1;
288 SvFLAGS(sv) = 0;
fd0854ff 289 sv->sv_debug_optype = PL_op ? PL_op->op_type : 0;
e385c3bf
DM
290 sv->sv_debug_line = (U16) (PL_parser && PL_parser->copline != NOLINE
291 ? PL_parser->copline
292 : PL_curcop
f24aceb1
DM
293 ? CopLINE(PL_curcop)
294 : 0
e385c3bf 295 );
fd0854ff
DM
296 sv->sv_debug_inpad = 0;
297 sv->sv_debug_cloned = 0;
fd0854ff 298 sv->sv_debug_file = PL_curcop ? savepv(CopFILE(PL_curcop)): NULL;
d7a2c63c
MHM
299
300 sv->sv_debug_serial = PL_sv_serial++;
301
302 MEM_LOG_NEW_SV(sv, file, line, func);
303 DEBUG_m(PerlIO_printf(Perl_debug_log, "0x%"UVxf": (%05ld) new_SV (from %s:%d [%s])\n",
304 PTR2UV(sv), (long)sv->sv_debug_serial, file, line, func));
305
eba0f806
DM
306 return sv;
307}
d7a2c63c 308# define new_SV(p) (p)=S_new_SV(aTHX_ __FILE__, __LINE__, FUNCTION__)
eba0f806
DM
309
310#else
311# define new_SV(p) \
053fc874 312 STMT_START { \
053fc874
GS
313 if (PL_sv_root) \
314 uproot_SV(p); \
315 else \
cac9b346 316 (p) = S_more_sv(aTHX); \
053fc874
GS
317 SvANY(p) = 0; \
318 SvREFCNT(p) = 1; \
319 SvFLAGS(p) = 0; \
d7a2c63c 320 MEM_LOG_NEW_SV(p, __FILE__, __LINE__, FUNCTION__); \
053fc874 321 } STMT_END
eba0f806 322#endif
463ee0b2 323
645c22ef
DM
324
325/* del_SV(): return an empty SV head to the free list */
326
a0d0e21e 327#ifdef DEBUGGING
4561caa4 328
053fc874
GS
329#define del_SV(p) \
330 STMT_START { \
aea4f609 331 if (DEBUG_D_TEST) \
053fc874
GS
332 del_sv(p); \
333 else \
334 plant_SV(p); \
053fc874 335 } STMT_END
a0d0e21e 336
76e3520e 337STATIC void
cea2e8a9 338S_del_sv(pTHX_ SV *p)
463ee0b2 339{
97aff369 340 dVAR;
7918f24d
NC
341
342 PERL_ARGS_ASSERT_DEL_SV;
343
aea4f609 344 if (DEBUG_D_TEST) {
4633a7c4 345 SV* sva;
a3b680e6 346 bool ok = 0;
daba3364 347 for (sva = PL_sv_arenaroot; sva; sva = MUTABLE_SV(SvANY(sva))) {
53c1dcc0
AL
348 const SV * const sv = sva + 1;
349 const SV * const svend = &sva[SvREFCNT(sva)];
c0ff570e 350 if (p >= sv && p < svend) {
a0d0e21e 351 ok = 1;
c0ff570e
NC
352 break;
353 }
a0d0e21e
LW
354 }
355 if (!ok) {
9b387841
NC
356 Perl_ck_warner_d(aTHX_ packWARN(WARN_INTERNAL),
357 "Attempt to free non-arena SV: 0x%"UVxf
358 pTHX__FORMAT, PTR2UV(p) pTHX__VALUE);
a0d0e21e
LW
359 return;
360 }
361 }
4561caa4 362 plant_SV(p);
463ee0b2 363}
a0d0e21e 364
4561caa4
CS
365#else /* ! DEBUGGING */
366
367#define del_SV(p) plant_SV(p)
368
369#endif /* DEBUGGING */
463ee0b2 370
645c22ef
DM
371
372/*
ccfc67b7
JH
373=head1 SV Manipulation Functions
374
645c22ef
DM
375=for apidoc sv_add_arena
376
377Given a chunk of memory, link it to the head of the list of arenas,
378and split it into a list of free SVs.
379
380=cut
381*/
382
d2bd4e7f
NC
383static void
384S_sv_add_arena(pTHX_ char *const ptr, const U32 size, const U32 flags)
463ee0b2 385{
97aff369 386 dVAR;
daba3364 387 SV *const sva = MUTABLE_SV(ptr);
463ee0b2
LW
388 register SV* sv;
389 register SV* svend;
4633a7c4 390
7918f24d
NC
391 PERL_ARGS_ASSERT_SV_ADD_ARENA;
392
4633a7c4 393 /* The first SV in an arena isn't an SV. */
3280af22 394 SvANY(sva) = (void *) PL_sv_arenaroot; /* ptr to next arena */
4633a7c4
LW
395 SvREFCNT(sva) = size / sizeof(SV); /* number of SV slots */
396 SvFLAGS(sva) = flags; /* FAKE if not to be freed */
397
3280af22
NIS
398 PL_sv_arenaroot = sva;
399 PL_sv_root = sva + 1;
4633a7c4
LW
400
401 svend = &sva[SvREFCNT(sva) - 1];
402 sv = sva + 1;
463ee0b2 403 while (sv < svend) {
3eef1deb 404 SvARENA_CHAIN_SET(sv, (sv + 1));
03e36789 405#ifdef DEBUGGING
978b032e 406 SvREFCNT(sv) = 0;
03e36789 407#endif
4b69cbe3 408 /* Must always set typemask because it's always checked in on cleanup
03e36789 409 when the arenas are walked looking for objects. */
8990e307 410 SvFLAGS(sv) = SVTYPEMASK;
463ee0b2
LW
411 sv++;
412 }
3eef1deb 413 SvARENA_CHAIN_SET(sv, 0);
03e36789
NC
414#ifdef DEBUGGING
415 SvREFCNT(sv) = 0;
416#endif
4633a7c4
LW
417 SvFLAGS(sv) = SVTYPEMASK;
418}
419
055972dc
DM
420/* visit(): call the named function for each non-free SV in the arenas
421 * whose flags field matches the flags/mask args. */
645c22ef 422
5226ed68 423STATIC I32
de37a194 424S_visit(pTHX_ SVFUNC_t f, const U32 flags, const U32 mask)
8990e307 425{
97aff369 426 dVAR;
4633a7c4 427 SV* sva;
5226ed68 428 I32 visited = 0;
8990e307 429
7918f24d
NC
430 PERL_ARGS_ASSERT_VISIT;
431
daba3364 432 for (sva = PL_sv_arenaroot; sva; sva = MUTABLE_SV(SvANY(sva))) {
53c1dcc0 433 register const SV * const svend = &sva[SvREFCNT(sva)];
a3b680e6 434 register SV* sv;
4561caa4 435 for (sv = sva + 1; sv < svend; ++sv) {
055972dc
DM
436 if (SvTYPE(sv) != SVTYPEMASK
437 && (sv->sv_flags & mask) == flags
438 && SvREFCNT(sv))
439 {
acfe0abc 440 (FCALL)(aTHX_ sv);
5226ed68
JH
441 ++visited;
442 }
8990e307
LW
443 }
444 }
5226ed68 445 return visited;
8990e307
LW
446}
447
758a08c3
JH
448#ifdef DEBUGGING
449
645c22ef
DM
450/* called by sv_report_used() for each live SV */
451
452static void
5fa45a31 453do_report_used(pTHX_ SV *const sv)
645c22ef
DM
454{
455 if (SvTYPE(sv) != SVTYPEMASK) {
456 PerlIO_printf(Perl_debug_log, "****\n");
457 sv_dump(sv);
458 }
459}
758a08c3 460#endif
645c22ef
DM
461
462/*
463=for apidoc sv_report_used
464
465Dump the contents of all SVs not yet freed. (Debugging aid).
466
467=cut
468*/
469
8990e307 470void
864dbfa3 471Perl_sv_report_used(pTHX)
4561caa4 472{
ff270d3a 473#ifdef DEBUGGING
055972dc 474 visit(do_report_used, 0, 0);
96a5add6
AL
475#else
476 PERL_UNUSED_CONTEXT;
ff270d3a 477#endif
4561caa4
CS
478}
479
645c22ef
DM
480/* called by sv_clean_objs() for each live SV */
481
482static void
de37a194 483do_clean_objs(pTHX_ SV *const ref)
645c22ef 484{
97aff369 485 dVAR;
ea724faa
NC
486 assert (SvROK(ref));
487 {
823a54a3
AL
488 SV * const target = SvRV(ref);
489 if (SvOBJECT(target)) {
490 DEBUG_D((PerlIO_printf(Perl_debug_log, "Cleaning object ref:\n "), sv_dump(ref)));
491 if (SvWEAKREF(ref)) {
492 sv_del_backref(target, ref);
493 SvWEAKREF_off(ref);
494 SvRV_set(ref, NULL);
495 } else {
496 SvROK_off(ref);
497 SvRV_set(ref, NULL);
498 SvREFCNT_dec(target);
499 }
645c22ef
DM
500 }
501 }
502
503 /* XXX Might want to check arrays, etc. */
504}
505
506/* called by sv_clean_objs() for each live SV */
507
508#ifndef DISABLE_DESTRUCTOR_KLUDGE
509static void
f30de749 510do_clean_named_objs(pTHX_ SV *const sv)
645c22ef 511{
97aff369 512 dVAR;
ea724faa 513 assert(SvTYPE(sv) == SVt_PVGV);
d011219a
NC
514 assert(isGV_with_GP(sv));
515 if (GvGP(sv)) {
c69033f2
NC
516 if ((
517#ifdef PERL_DONT_CREATE_GVSV
518 GvSV(sv) &&
519#endif
520 SvOBJECT(GvSV(sv))) ||
645c22ef
DM
521 (GvAV(sv) && SvOBJECT(GvAV(sv))) ||
522 (GvHV(sv) && SvOBJECT(GvHV(sv))) ||
9c12f1e5
RGS
523 /* In certain rare cases GvIOp(sv) can be NULL, which would make SvOBJECT(GvIO(sv)) dereference NULL. */
524 (GvIO(sv) ? (SvFLAGS(GvIOp(sv)) & SVs_OBJECT) : 0) ||
645c22ef
DM
525 (GvCV(sv) && SvOBJECT(GvCV(sv))) )
526 {
527 DEBUG_D((PerlIO_printf(Perl_debug_log, "Cleaning named glob object:\n "), sv_dump(sv)));
ec5f3c78 528 SvFLAGS(sv) |= SVf_BREAK;
645c22ef
DM
529 SvREFCNT_dec(sv);
530 }
531 }
532}
533#endif
534
535/*
536=for apidoc sv_clean_objs
537
538Attempt to destroy all objects not yet freed
539
540=cut
541*/
542
4561caa4 543void
864dbfa3 544Perl_sv_clean_objs(pTHX)
4561caa4 545{
97aff369 546 dVAR;
3280af22 547 PL_in_clean_objs = TRUE;
055972dc 548 visit(do_clean_objs, SVf_ROK, SVf_ROK);
4561caa4 549#ifndef DISABLE_DESTRUCTOR_KLUDGE
2d0f3c12 550 /* some barnacles may yet remain, clinging to typeglobs */
d011219a 551 visit(do_clean_named_objs, SVt_PVGV|SVpgv_GP, SVTYPEMASK|SVp_POK|SVpgv_GP);
4561caa4 552#endif
3280af22 553 PL_in_clean_objs = FALSE;
4561caa4
CS
554}
555
645c22ef
DM
556/* called by sv_clean_all() for each live SV */
557
558static void
de37a194 559do_clean_all(pTHX_ SV *const sv)
645c22ef 560{
97aff369 561 dVAR;
daba3364 562 if (sv == (const SV *) PL_fdpid || sv == (const SV *)PL_strtab) {
cddfcddc 563 /* don't clean pid table and strtab */
d17ea597 564 return;
cddfcddc 565 }
645c22ef
DM
566 DEBUG_D((PerlIO_printf(Perl_debug_log, "Cleaning loops: SV at 0x%"UVxf"\n", PTR2UV(sv)) ));
567 SvFLAGS(sv) |= SVf_BREAK;
568 SvREFCNT_dec(sv);
569}
570
571/*
572=for apidoc sv_clean_all
573
574Decrement the refcnt of each remaining SV, possibly triggering a
575cleanup. This function may have to be called multiple times to free
ff276b08 576SVs which are in complex self-referential hierarchies.
645c22ef
DM
577
578=cut
579*/
580
5226ed68 581I32
864dbfa3 582Perl_sv_clean_all(pTHX)
8990e307 583{
97aff369 584 dVAR;
5226ed68 585 I32 cleaned;
3280af22 586 PL_in_clean_all = TRUE;
055972dc 587 cleaned = visit(do_clean_all, 0,0);
3280af22 588 PL_in_clean_all = FALSE;
5226ed68 589 return cleaned;
8990e307 590}
463ee0b2 591
5e258f8c
JC
592/*
593 ARENASETS: a meta-arena implementation which separates arena-info
594 into struct arena_set, which contains an array of struct
595 arena_descs, each holding info for a single arena. By separating
596 the meta-info from the arena, we recover the 1st slot, formerly
597 borrowed for list management. The arena_set is about the size of an
39244528 598 arena, avoiding the needless malloc overhead of a naive linked-list.
5e258f8c
JC
599
600 The cost is 1 arena-set malloc per ~320 arena-mallocs, + the unused
601 memory in the last arena-set (1/2 on average). In trade, we get
602 back the 1st slot in each arena (ie 1.7% of a CV-arena, less for
d2a0f284 603 smaller types). The recovery of the wasted space allows use of
e15dad31
JC
604 small arenas for large, rare body types, by changing array* fields
605 in body_details_by_type[] below.
5e258f8c 606*/
5e258f8c 607struct arena_desc {
398c677b
NC
608 char *arena; /* the raw storage, allocated aligned */
609 size_t size; /* its size ~4k typ */
e5973ed5 610 svtype utype; /* bodytype stored in arena */
5e258f8c
JC
611};
612
e6148039
NC
613struct arena_set;
614
615/* Get the maximum number of elements in set[] such that struct arena_set
e15dad31 616 will fit within PERL_ARENA_SIZE, which is probably just under 4K, and
e6148039
NC
617 therefore likely to be 1 aligned memory page. */
618
619#define ARENAS_PER_SET ((PERL_ARENA_SIZE - sizeof(struct arena_set*) \
620 - 2 * sizeof(int)) / sizeof (struct arena_desc))
5e258f8c
JC
621
622struct arena_set {
623 struct arena_set* next;
0a848332
NC
624 unsigned int set_size; /* ie ARENAS_PER_SET */
625 unsigned int curr; /* index of next available arena-desc */
5e258f8c
JC
626 struct arena_desc set[ARENAS_PER_SET];
627};
628
645c22ef
DM
629/*
630=for apidoc sv_free_arenas
631
632Deallocate the memory used by all arenas. Note that all the individual SV
633heads and bodies within the arenas must already have been freed.
634
635=cut
636*/
4633a7c4 637void
864dbfa3 638Perl_sv_free_arenas(pTHX)
4633a7c4 639{
97aff369 640 dVAR;
4633a7c4
LW
641 SV* sva;
642 SV* svanext;
0a848332 643 unsigned int i;
4633a7c4
LW
644
645 /* Free arenas here, but be careful about fake ones. (We assume
646 contiguity of the fake ones with the corresponding real ones.) */
647
3280af22 648 for (sva = PL_sv_arenaroot; sva; sva = svanext) {
daba3364 649 svanext = MUTABLE_SV(SvANY(sva));
4633a7c4 650 while (svanext && SvFAKE(svanext))
daba3364 651 svanext = MUTABLE_SV(SvANY(svanext));
4633a7c4
LW
652
653 if (!SvFAKE(sva))
1df70142 654 Safefree(sva);
4633a7c4 655 }
93e68bfb 656
5e258f8c 657 {
0a848332
NC
658 struct arena_set *aroot = (struct arena_set*) PL_body_arenas;
659
660 while (aroot) {
661 struct arena_set *current = aroot;
662 i = aroot->curr;
663 while (i--) {
5e258f8c
JC
664 assert(aroot->set[i].arena);
665 Safefree(aroot->set[i].arena);
666 }
0a848332
NC
667 aroot = aroot->next;
668 Safefree(current);
5e258f8c
JC
669 }
670 }
dc8220bf 671 PL_body_arenas = 0;
fdda85ca 672
0a848332
NC
673 i = PERL_ARENA_ROOTS_SIZE;
674 while (i--)
93e68bfb 675 PL_body_roots[i] = 0;
93e68bfb 676
43c5f42d 677 Safefree(PL_nice_chunk);
bd61b366 678 PL_nice_chunk = NULL;
3280af22
NIS
679 PL_nice_chunk_size = 0;
680 PL_sv_arenaroot = 0;
681 PL_sv_root = 0;
4633a7c4
LW
682}
683
bd81e77b
NC
684/*
685 Here are mid-level routines that manage the allocation of bodies out
686 of the various arenas. There are 5 kinds of arenas:
29489e7c 687
bd81e77b
NC
688 1. SV-head arenas, which are discussed and handled above
689 2. regular body arenas
690 3. arenas for reduced-size bodies
691 4. Hash-Entry arenas
692 5. pte arenas (thread related)
29489e7c 693
bd81e77b
NC
694 Arena types 2 & 3 are chained by body-type off an array of
695 arena-root pointers, which is indexed by svtype. Some of the
696 larger/less used body types are malloced singly, since a large
697 unused block of them is wasteful. Also, several svtypes dont have
698 bodies; the data fits into the sv-head itself. The arena-root
699 pointer thus has a few unused root-pointers (which may be hijacked
700 later for arena types 4,5)
29489e7c 701
bd81e77b
NC
702 3 differs from 2 as an optimization; some body types have several
703 unused fields in the front of the structure (which are kept in-place
704 for consistency). These bodies can be allocated in smaller chunks,
705 because the leading fields arent accessed. Pointers to such bodies
706 are decremented to point at the unused 'ghost' memory, knowing that
707 the pointers are used with offsets to the real memory.
29489e7c 708
bd81e77b
NC
709 HE, HEK arenas are managed separately, with separate code, but may
710 be merge-able later..
711
712 PTE arenas are not sv-bodies, but they share these mid-level
713 mechanics, so are considered here. The new mid-level mechanics rely
714 on the sv_type of the body being allocated, so we just reserve one
715 of the unused body-slots for PTEs, then use it in those (2) PTE
716 contexts below (line ~10k)
717*/
718
bd26d9a3 719/* get_arena(size): this creates custom-sized arenas
5e258f8c
JC
720 TBD: export properly for hv.c: S_more_he().
721*/
722void*
e5973ed5 723Perl_get_arena(pTHX_ const size_t arena_size, const svtype bodytype)
5e258f8c 724{
7a89be66 725 dVAR;
5e258f8c 726 struct arena_desc* adesc;
39244528 727 struct arena_set *aroot = (struct arena_set*) PL_body_arenas;
0a848332 728 unsigned int curr;
5e258f8c 729
476a1e16
JC
730 /* shouldnt need this
731 if (!arena_size) arena_size = PERL_ARENA_SIZE;
732 */
5e258f8c
JC
733
734 /* may need new arena-set to hold new arena */
39244528
NC
735 if (!aroot || aroot->curr >= aroot->set_size) {
736 struct arena_set *newroot;
5e258f8c
JC
737 Newxz(newroot, 1, struct arena_set);
738 newroot->set_size = ARENAS_PER_SET;
39244528
NC
739 newroot->next = aroot;
740 aroot = newroot;
741 PL_body_arenas = (void *) newroot;
52944de8 742 DEBUG_m(PerlIO_printf(Perl_debug_log, "new arenaset %p\n", (void*)aroot));
5e258f8c
JC
743 }
744
745 /* ok, now have arena-set with at least 1 empty/available arena-desc */
39244528
NC
746 curr = aroot->curr++;
747 adesc = &(aroot->set[curr]);
5e258f8c
JC
748 assert(!adesc->arena);
749
89086707 750 Newx(adesc->arena, arena_size, char);
5e258f8c 751 adesc->size = arena_size;
e5973ed5 752 adesc->utype = bodytype;
d67b3c53
JH
753 DEBUG_m(PerlIO_printf(Perl_debug_log, "arena %d added: %p size %"UVuf"\n",
754 curr, (void*)adesc->arena, (UV)arena_size));
5e258f8c
JC
755
756 return adesc->arena;
5e258f8c
JC
757}
758
53c1dcc0 759
bd81e77b 760/* return a thing to the free list */
29489e7c 761
bd81e77b
NC
762#define del_body(thing, root) \
763 STMT_START { \
00b6aa41 764 void ** const thing_copy = (void **)thing;\
bd81e77b
NC
765 *thing_copy = *root; \
766 *root = (void*)thing_copy; \
bd81e77b 767 } STMT_END
29489e7c 768
bd81e77b 769/*
d2a0f284
JC
770
771=head1 SV-Body Allocation
772
773Allocation of SV-bodies is similar to SV-heads, differing as follows;
774the allocation mechanism is used for many body types, so is somewhat
775more complicated, it uses arena-sets, and has no need for still-live
776SV detection.
777
778At the outermost level, (new|del)_X*V macros return bodies of the
779appropriate type. These macros call either (new|del)_body_type or
780(new|del)_body_allocated macro pairs, depending on specifics of the
781type. Most body types use the former pair, the latter pair is used to
782allocate body types with "ghost fields".
783
784"ghost fields" are fields that are unused in certain types, and
69ba284b 785consequently don't need to actually exist. They are declared because
d2a0f284
JC
786they're part of a "base type", which allows use of functions as
787methods. The simplest examples are AVs and HVs, 2 aggregate types
788which don't use the fields which support SCALAR semantics.
789
69ba284b 790For these types, the arenas are carved up into appropriately sized
d2a0f284
JC
791chunks, we thus avoid wasted memory for those unaccessed members.
792When bodies are allocated, we adjust the pointer back in memory by the
69ba284b 793size of the part not allocated, so it's as if we allocated the full
d2a0f284
JC
794structure. (But things will all go boom if you write to the part that
795is "not there", because you'll be overwriting the last members of the
796preceding structure in memory.)
797
69ba284b
NC
798We calculate the correction using the STRUCT_OFFSET macro on the first
799member present. If the allocated structure is smaller (no initial NV
800actually allocated) then the net effect is to subtract the size of the NV
801from the pointer, to return a new pointer as if an initial NV were actually
802allocated. (We were using structures named *_allocated for this, but
803this turned out to be a subtle bug, because a structure without an NV
804could have a lower alignment constraint, but the compiler is allowed to
805optimised accesses based on the alignment constraint of the actual pointer
806to the full structure, for example, using a single 64 bit load instruction
807because it "knows" that two adjacent 32 bit members will be 8-byte aligned.)
d2a0f284
JC
808
809This is the same trick as was used for NV and IV bodies. Ironically it
810doesn't need to be used for NV bodies any more, because NV is now at
811the start of the structure. IV bodies don't need it either, because
812they are no longer allocated.
813
814In turn, the new_body_* allocators call S_new_body(), which invokes
815new_body_inline macro, which takes a lock, and takes a body off the
816linked list at PL_body_roots[sv_type], calling S_more_bodies() if
817necessary to refresh an empty list. Then the lock is released, and
818the body is returned.
819
820S_more_bodies calls get_arena(), and carves it up into an array of N
821bodies, which it strings into a linked list. It looks up arena-size
822and body-size from the body_details table described below, thus
823supporting the multiple body-types.
824
825If PURIFY is defined, or PERL_ARENA_SIZE=0, arenas are not used, and
826the (new|del)_X*V macros are mapped directly to malloc/free.
827
828*/
829
830/*
831
832For each sv-type, struct body_details bodies_by_type[] carries
833parameters which control these aspects of SV handling:
834
835Arena_size determines whether arenas are used for this body type, and if
836so, how big they are. PURIFY or PERL_ARENA_SIZE=0 set this field to
837zero, forcing individual mallocs and frees.
838
839Body_size determines how big a body is, and therefore how many fit into
840each arena. Offset carries the body-pointer adjustment needed for
69ba284b 841"ghost fields", and is used in *_allocated macros.
d2a0f284
JC
842
843But its main purpose is to parameterize info needed in
844Perl_sv_upgrade(). The info here dramatically simplifies the function
69ba284b 845vs the implementation in 5.8.8, making it table-driven. All fields
d2a0f284
JC
846are used for this, except for arena_size.
847
848For the sv-types that have no bodies, arenas are not used, so those
849PL_body_roots[sv_type] are unused, and can be overloaded. In
850something of a special case, SVt_NULL is borrowed for HE arenas;
c6f8b1d0 851PL_body_roots[HE_SVSLOT=SVt_NULL] is filled by S_more_he, but the
d2a0f284 852bodies_by_type[SVt_NULL] slot is not used, as the table is not
c6f8b1d0 853available in hv.c.
d2a0f284 854
c6f8b1d0
JC
855PTEs also use arenas, but are never seen in Perl_sv_upgrade. Nonetheless,
856they get their own slot in bodies_by_type[PTE_SVSLOT =SVt_IV], so they can
857just use the same allocation semantics. At first, PTEs were also
858overloaded to a non-body sv-type, but this yielded hard-to-find malloc
859bugs, so was simplified by claiming a new slot. This choice has no
860consequence at this time.
d2a0f284 861
29489e7c
DM
862*/
863
bd81e77b 864struct body_details {
0fb58b32 865 U8 body_size; /* Size to allocate */
10666ae3 866 U8 copy; /* Size of structure to copy (may be shorter) */
0fb58b32 867 U8 offset;
10666ae3
NC
868 unsigned int type : 4; /* We have space for a sanity check. */
869 unsigned int cant_upgrade : 1; /* Cannot upgrade this type */
870 unsigned int zero_nv : 1; /* zero the NV when upgrading from this */
871 unsigned int arena : 1; /* Allocated from an arena */
872 size_t arena_size; /* Size of arena to allocate */
bd81e77b 873};
29489e7c 874
bd81e77b
NC
875#define HADNV FALSE
876#define NONV TRUE
29489e7c 877
d2a0f284 878
bd81e77b
NC
879#ifdef PURIFY
880/* With -DPURFIY we allocate everything directly, and don't use arenas.
881 This seems a rather elegant way to simplify some of the code below. */
882#define HASARENA FALSE
883#else
884#define HASARENA TRUE
885#endif
886#define NOARENA FALSE
29489e7c 887
d2a0f284
JC
888/* Size the arenas to exactly fit a given number of bodies. A count
889 of 0 fits the max number bodies into a PERL_ARENA_SIZE.block,
890 simplifying the default. If count > 0, the arena is sized to fit
891 only that many bodies, allowing arenas to be used for large, rare
892 bodies (XPVFM, XPVIO) without undue waste. The arena size is
893 limited by PERL_ARENA_SIZE, so we can safely oversize the
894 declarations.
895 */
95db5f15
MB
896#define FIT_ARENA0(body_size) \
897 ((size_t)(PERL_ARENA_SIZE / body_size) * body_size)
898#define FIT_ARENAn(count,body_size) \
899 ( count * body_size <= PERL_ARENA_SIZE) \
900 ? count * body_size \
901 : FIT_ARENA0 (body_size)
902#define FIT_ARENA(count,body_size) \
903 count \
904 ? FIT_ARENAn (count, body_size) \
905 : FIT_ARENA0 (body_size)
d2a0f284 906
bd81e77b
NC
907/* Calculate the length to copy. Specifically work out the length less any
908 final padding the compiler needed to add. See the comment in sv_upgrade
909 for why copying the padding proved to be a bug. */
29489e7c 910
bd81e77b
NC
911#define copy_length(type, last_member) \
912 STRUCT_OFFSET(type, last_member) \
daba3364 913 + sizeof (((type*)SvANY((const SV *)0))->last_member)
29489e7c 914
bd81e77b 915static const struct body_details bodies_by_type[] = {
10666ae3
NC
916 { sizeof(HE), 0, 0, SVt_NULL,
917 FALSE, NONV, NOARENA, FIT_ARENA(0, sizeof(HE)) },
d2a0f284 918
1cb9cd50 919 /* The bind placeholder pretends to be an RV for now.
c6f8b1d0 920 Also it's marked as "can't upgrade" to stop anyone using it before it's
1cb9cd50
NC
921 implemented. */
922 { 0, 0, 0, SVt_BIND, TRUE, NONV, NOARENA, 0 },
923
d2a0f284
JC
924 /* IVs are in the head, so the allocation size is 0.
925 However, the slot is overloaded for PTEs. */
926 { sizeof(struct ptr_tbl_ent), /* This is used for PTEs. */
927 sizeof(IV), /* This is used to copy out the IV body. */
10666ae3 928 STRUCT_OFFSET(XPVIV, xiv_iv), SVt_IV, FALSE, NONV,
d2a0f284
JC
929 NOARENA /* IVS don't need an arena */,
930 /* But PTEs need to know the size of their arena */
931 FIT_ARENA(0, sizeof(struct ptr_tbl_ent))
932 },
933
bd81e77b 934 /* 8 bytes on most ILP32 with IEEE doubles */
10666ae3 935 { sizeof(NV), sizeof(NV), 0, SVt_NV, FALSE, HADNV, HASARENA,
d2a0f284
JC
936 FIT_ARENA(0, sizeof(NV)) },
937
bd81e77b 938 /* 8 bytes on most ILP32 with IEEE doubles */
69ba284b
NC
939 { sizeof(XPV) - STRUCT_OFFSET(XPV, xpv_cur),
940 copy_length(XPV, xpv_len) - STRUCT_OFFSET(XPV, xpv_cur),
941 + STRUCT_OFFSET(XPV, xpv_cur),
942 SVt_PV, FALSE, NONV, HASARENA,
943 FIT_ARENA(0, sizeof(XPV) - STRUCT_OFFSET(XPV, xpv_cur)) },
d2a0f284 944
bd81e77b 945 /* 12 */
69ba284b
NC
946 { sizeof(XPVIV) - STRUCT_OFFSET(XPV, xpv_cur),
947 copy_length(XPVIV, xiv_u) - STRUCT_OFFSET(XPV, xpv_cur),
948 + STRUCT_OFFSET(XPVIV, xpv_cur),
949 SVt_PVIV, FALSE, NONV, HASARENA,
950 FIT_ARENA(0, sizeof(XPV) - STRUCT_OFFSET(XPV, xpv_cur)) },
d2a0f284 951
bd81e77b 952 /* 20 */
10666ae3 953 { sizeof(XPVNV), copy_length(XPVNV, xiv_u), 0, SVt_PVNV, FALSE, HADNV,
d2a0f284
JC
954 HASARENA, FIT_ARENA(0, sizeof(XPVNV)) },
955
bd81e77b 956 /* 28 */
10666ae3 957 { sizeof(XPVMG), copy_length(XPVMG, xmg_stash), 0, SVt_PVMG, FALSE, HADNV,
d2a0f284 958 HASARENA, FIT_ARENA(0, sizeof(XPVMG)) },
4df7f6af 959
288b8c02 960 /* something big */
b6f60916
NC
961 { sizeof(regexp) - STRUCT_OFFSET(regexp, xpv_cur),
962 sizeof(regexp) - STRUCT_OFFSET(regexp, xpv_cur),
963 + STRUCT_OFFSET(regexp, xpv_cur),
08e44740 964 SVt_REGEXP, FALSE, NONV, HASARENA,
b6f60916 965 FIT_ARENA(0, sizeof(regexp) - STRUCT_OFFSET(regexp, xpv_cur))
5c35adbb 966 },
4df7f6af 967
bd81e77b 968 /* 48 */
10666ae3 969 { sizeof(XPVGV), sizeof(XPVGV), 0, SVt_PVGV, TRUE, HADNV,
d2a0f284
JC
970 HASARENA, FIT_ARENA(0, sizeof(XPVGV)) },
971
bd81e77b 972 /* 64 */
10666ae3 973 { sizeof(XPVLV), sizeof(XPVLV), 0, SVt_PVLV, TRUE, HADNV,
d2a0f284
JC
974 HASARENA, FIT_ARENA(0, sizeof(XPVLV)) },
975
69ba284b
NC
976 { sizeof(XPVAV) - STRUCT_OFFSET(XPVAV, xav_fill),
977 copy_length(XPVAV, xmg_stash) - STRUCT_OFFSET(XPVAV, xav_fill),
978 + STRUCT_OFFSET(XPVAV, xav_fill),
979 SVt_PVAV, TRUE, NONV, HASARENA,
980 FIT_ARENA(0, sizeof(XPVAV) - STRUCT_OFFSET(XPVAV, xav_fill)) },
d2a0f284 981
69ba284b
NC
982 { sizeof(XPVHV) - STRUCT_OFFSET(XPVHV, xhv_fill),
983 copy_length(XPVHV, xmg_stash) - STRUCT_OFFSET(XPVHV, xhv_fill),
984 + STRUCT_OFFSET(XPVHV, xhv_fill),
985 SVt_PVHV, TRUE, NONV, HASARENA,
986 FIT_ARENA(0, sizeof(XPVHV) - STRUCT_OFFSET(XPVHV, xhv_fill)) },
d2a0f284 987
c84c4652 988 /* 56 */
69ba284b
NC
989 { sizeof(XPVCV) - STRUCT_OFFSET(XPVCV, xpv_cur),
990 sizeof(XPVCV) - STRUCT_OFFSET(XPVCV, xpv_cur),
991 + STRUCT_OFFSET(XPVCV, xpv_cur),
992 SVt_PVCV, TRUE, NONV, HASARENA,
993 FIT_ARENA(0, sizeof(XPVCV) - STRUCT_OFFSET(XPVCV, xpv_cur)) },
994
995 { sizeof(XPVFM) - STRUCT_OFFSET(XPVFM, xpv_cur),
996 sizeof(XPVFM) - STRUCT_OFFSET(XPVFM, xpv_cur),
997 + STRUCT_OFFSET(XPVFM, xpv_cur),
998 SVt_PVFM, TRUE, NONV, NOARENA,
999 FIT_ARENA(20, sizeof(XPVFM) - STRUCT_OFFSET(XPVFM, xpv_cur)) },
d2a0f284
JC
1000
1001 /* XPVIO is 84 bytes, fits 48x */
b6f60916
NC
1002 { sizeof(XPVIO) - STRUCT_OFFSET(XPVIO, xpv_cur),
1003 sizeof(XPVIO) - STRUCT_OFFSET(XPVIO, xpv_cur),
1004 + STRUCT_OFFSET(XPVIO, xpv_cur),
1005 SVt_PVIO, TRUE, NONV, HASARENA,
1006 FIT_ARENA(24, sizeof(XPVIO) - STRUCT_OFFSET(XPVIO, xpv_cur)) },
bd81e77b 1007};
29489e7c 1008
d2a0f284
JC
1009#define new_body_type(sv_type) \
1010 (void *)((char *)S_new_body(aTHX_ sv_type))
29489e7c 1011
bd81e77b
NC
1012#define del_body_type(p, sv_type) \
1013 del_body(p, &PL_body_roots[sv_type])
29489e7c 1014
29489e7c 1015
bd81e77b 1016#define new_body_allocated(sv_type) \
d2a0f284 1017 (void *)((char *)S_new_body(aTHX_ sv_type) \
bd81e77b 1018 - bodies_by_type[sv_type].offset)
29489e7c 1019
bd81e77b
NC
1020#define del_body_allocated(p, sv_type) \
1021 del_body(p + bodies_by_type[sv_type].offset, &PL_body_roots[sv_type])
29489e7c 1022
29489e7c 1023
bd81e77b
NC
1024#define my_safemalloc(s) (void*)safemalloc(s)
1025#define my_safecalloc(s) (void*)safecalloc(s, 1)
1026#define my_safefree(p) safefree((char*)p)
29489e7c 1027
bd81e77b 1028#ifdef PURIFY
29489e7c 1029
bd81e77b
NC
1030#define new_XNV() my_safemalloc(sizeof(XPVNV))
1031#define del_XNV(p) my_safefree(p)
29489e7c 1032
bd81e77b
NC
1033#define new_XPVNV() my_safemalloc(sizeof(XPVNV))
1034#define del_XPVNV(p) my_safefree(p)
29489e7c 1035
bd81e77b
NC
1036#define new_XPVAV() my_safemalloc(sizeof(XPVAV))
1037#define del_XPVAV(p) my_safefree(p)
29489e7c 1038
bd81e77b
NC
1039#define new_XPVHV() my_safemalloc(sizeof(XPVHV))
1040#define del_XPVHV(p) my_safefree(p)
29489e7c 1041
bd81e77b
NC
1042#define new_XPVMG() my_safemalloc(sizeof(XPVMG))
1043#define del_XPVMG(p) my_safefree(p)
29489e7c 1044
bd81e77b
NC
1045#define new_XPVGV() my_safemalloc(sizeof(XPVGV))
1046#define del_XPVGV(p) my_safefree(p)
29489e7c 1047
bd81e77b 1048#else /* !PURIFY */
29489e7c 1049
bd81e77b
NC
1050#define new_XNV() new_body_type(SVt_NV)
1051#define del_XNV(p) del_body_type(p, SVt_NV)
29489e7c 1052
bd81e77b
NC
1053#define new_XPVNV() new_body_type(SVt_PVNV)
1054#define del_XPVNV(p) del_body_type(p, SVt_PVNV)
29489e7c 1055
bd81e77b
NC
1056#define new_XPVAV() new_body_allocated(SVt_PVAV)
1057#define del_XPVAV(p) del_body_allocated(p, SVt_PVAV)
645c22ef 1058
bd81e77b
NC
1059#define new_XPVHV() new_body_allocated(SVt_PVHV)
1060#define del_XPVHV(p) del_body_allocated(p, SVt_PVHV)
645c22ef 1061
bd81e77b
NC
1062#define new_XPVMG() new_body_type(SVt_PVMG)
1063#define del_XPVMG(p) del_body_type(p, SVt_PVMG)
645c22ef 1064
bd81e77b
NC
1065#define new_XPVGV() new_body_type(SVt_PVGV)
1066#define del_XPVGV(p) del_body_type(p, SVt_PVGV)
1d7c1841 1067
bd81e77b 1068#endif /* PURIFY */
93e68bfb 1069
bd81e77b 1070/* no arena for you! */
93e68bfb 1071
bd81e77b 1072#define new_NOARENA(details) \
d2a0f284 1073 my_safemalloc((details)->body_size + (details)->offset)
bd81e77b 1074#define new_NOARENAZ(details) \
d2a0f284
JC
1075 my_safecalloc((details)->body_size + (details)->offset)
1076
1077STATIC void *
de37a194 1078S_more_bodies (pTHX_ const svtype sv_type)
d2a0f284
JC
1079{
1080 dVAR;
1081 void ** const root = &PL_body_roots[sv_type];
96a5add6 1082 const struct body_details * const bdp = &bodies_by_type[sv_type];
d2a0f284
JC
1083 const size_t body_size = bdp->body_size;
1084 char *start;
1085 const char *end;
d8fca402 1086 const size_t arena_size = Perl_malloc_good_size(bdp->arena_size);
0b2d3faa 1087#if defined(DEBUGGING) && !defined(PERL_GLOBAL_STRUCT_PRIVATE)
23e9d66c
NC
1088 static bool done_sanity_check;
1089
0b2d3faa
JH
1090 /* PERL_GLOBAL_STRUCT_PRIVATE cannot coexist with global
1091 * variables like done_sanity_check. */
10666ae3 1092 if (!done_sanity_check) {
ea471437 1093 unsigned int i = SVt_LAST;
10666ae3
NC
1094
1095 done_sanity_check = TRUE;
1096
1097 while (i--)
1098 assert (bodies_by_type[i].type == i);
1099 }
1100#endif
1101
23e9d66c
NC
1102 assert(bdp->arena_size);
1103
d8fca402 1104 start = (char*) Perl_get_arena(aTHX_ arena_size, sv_type);
d2a0f284 1105
d8fca402 1106 end = start + arena_size - 2 * body_size;
d2a0f284 1107
d2a0f284 1108 /* computed count doesnt reflect the 1st slot reservation */
d8fca402
NC
1109#if defined(MYMALLOC) || defined(HAS_MALLOC_GOOD_SIZE)
1110 DEBUG_m(PerlIO_printf(Perl_debug_log,
1111 "arena %p end %p arena-size %d (from %d) type %d "
1112 "size %d ct %d\n",
1113 (void*)start, (void*)end, (int)arena_size,
1114 (int)bdp->arena_size, sv_type, (int)body_size,
1115 (int)arena_size / (int)body_size));
1116#else
d2a0f284
JC
1117 DEBUG_m(PerlIO_printf(Perl_debug_log,
1118 "arena %p end %p arena-size %d type %d size %d ct %d\n",
6c9570dc 1119 (void*)start, (void*)end,
0e84aef4
JH
1120 (int)bdp->arena_size, sv_type, (int)body_size,
1121 (int)bdp->arena_size / (int)body_size));
d8fca402 1122#endif
d2a0f284
JC
1123 *root = (void *)start;
1124
d8fca402 1125 while (start <= end) {
d2a0f284
JC
1126 char * const next = start + body_size;
1127 *(void**) start = (void *)next;
1128 start = next;
1129 }
1130 *(void **)start = 0;
1131
1132 return *root;
1133}
1134
1135/* grab a new thing from the free list, allocating more if necessary.
1136 The inline version is used for speed in hot routines, and the
1137 function using it serves the rest (unless PURIFY).
1138*/
1139#define new_body_inline(xpv, sv_type) \
1140 STMT_START { \
1141 void ** const r3wt = &PL_body_roots[sv_type]; \
11b79775
DD
1142 xpv = (PTR_TBL_ENT_t*) (*((void **)(r3wt)) \
1143 ? *((void **)(r3wt)) : more_bodies(sv_type)); \
d2a0f284 1144 *(r3wt) = *(void**)(xpv); \
d2a0f284
JC
1145 } STMT_END
1146
1147#ifndef PURIFY
1148
1149STATIC void *
de37a194 1150S_new_body(pTHX_ const svtype sv_type)
d2a0f284
JC
1151{
1152 dVAR;
1153 void *xpv;
1154 new_body_inline(xpv, sv_type);
1155 return xpv;
1156}
1157
1158#endif
93e68bfb 1159
238b27b3
NC
1160static const struct body_details fake_rv =
1161 { 0, 0, 0, SVt_IV, FALSE, NONV, NOARENA, 0 };
1162
bd81e77b
NC
1163/*
1164=for apidoc sv_upgrade
93e68bfb 1165
bd81e77b
NC
1166Upgrade an SV to a more complex form. Generally adds a new body type to the
1167SV, then copies across as much information as possible from the old body.
1168You generally want to use the C<SvUPGRADE> macro wrapper. See also C<svtype>.
93e68bfb 1169
bd81e77b 1170=cut
93e68bfb 1171*/
93e68bfb 1172
bd81e77b 1173void
aad570aa 1174Perl_sv_upgrade(pTHX_ register SV *const sv, svtype new_type)
cac9b346 1175{
97aff369 1176 dVAR;
bd81e77b
NC
1177 void* old_body;
1178 void* new_body;
42d0e0b7 1179 const svtype old_type = SvTYPE(sv);
d2a0f284 1180 const struct body_details *new_type_details;
238b27b3 1181 const struct body_details *old_type_details
bd81e77b 1182 = bodies_by_type + old_type;
4df7f6af 1183 SV *referant = NULL;
cac9b346 1184
7918f24d
NC
1185 PERL_ARGS_ASSERT_SV_UPGRADE;
1186
1776cbe8
NC
1187 if (old_type == new_type)
1188 return;
1189
1190 /* This clause was purposefully added ahead of the early return above to
1191 the shared string hackery for (sort {$a <=> $b} keys %hash), with the
1192 inference by Nick I-S that it would fix other troublesome cases. See
1193 changes 7162, 7163 (f130fd4589cf5fbb24149cd4db4137c8326f49c1 and parent)
1194
1195 Given that shared hash key scalars are no longer PVIV, but PV, there is
1196 no longer need to unshare so as to free up the IVX slot for its proper
1197 purpose. So it's safe to move the early return earlier. */
1198
bd81e77b
NC
1199 if (new_type != SVt_PV && SvIsCOW(sv)) {
1200 sv_force_normal_flags(sv, 0);
1201 }
cac9b346 1202
bd81e77b 1203 old_body = SvANY(sv);
de042e1d 1204
bd81e77b
NC
1205 /* Copying structures onto other structures that have been neatly zeroed
1206 has a subtle gotcha. Consider XPVMG
cac9b346 1207
bd81e77b
NC
1208 +------+------+------+------+------+-------+-------+
1209 | NV | CUR | LEN | IV | MAGIC | STASH |
1210 +------+------+------+------+------+-------+-------+
1211 0 4 8 12 16 20 24 28
645c22ef 1212
bd81e77b
NC
1213 where NVs are aligned to 8 bytes, so that sizeof that structure is
1214 actually 32 bytes long, with 4 bytes of padding at the end:
08742458 1215
bd81e77b
NC
1216 +------+------+------+------+------+-------+-------+------+
1217 | NV | CUR | LEN | IV | MAGIC | STASH | ??? |
1218 +------+------+------+------+------+-------+-------+------+
1219 0 4 8 12 16 20 24 28 32
08742458 1220
bd81e77b 1221 so what happens if you allocate memory for this structure:
30f9da9e 1222
bd81e77b
NC
1223 +------+------+------+------+------+-------+-------+------+------+...
1224 | NV | CUR | LEN | IV | MAGIC | STASH | GP | NAME |
1225 +------+------+------+------+------+-------+-------+------+------+...
1226 0 4 8 12 16 20 24 28 32 36
bfc44f79 1227
bd81e77b
NC
1228 zero it, then copy sizeof(XPVMG) bytes on top of it? Not quite what you
1229 expect, because you copy the area marked ??? onto GP. Now, ??? may have
1230 started out as zero once, but it's quite possible that it isn't. So now,
1231 rather than a nicely zeroed GP, you have it pointing somewhere random.
1232 Bugs ensue.
bfc44f79 1233
bd81e77b
NC
1234 (In fact, GP ends up pointing at a previous GP structure, because the
1235 principle cause of the padding in XPVMG getting garbage is a copy of
6c9e42f7
NC
1236 sizeof(XPVMG) bytes from a XPVGV structure in sv_unglob. Right now
1237 this happens to be moot because XPVGV has been re-ordered, with GP
1238 no longer after STASH)
30f9da9e 1239
bd81e77b
NC
1240 So we are careful and work out the size of used parts of all the
1241 structures. */
bfc44f79 1242
bd81e77b
NC
1243 switch (old_type) {
1244 case SVt_NULL:
1245 break;
1246 case SVt_IV:
4df7f6af
NC
1247 if (SvROK(sv)) {
1248 referant = SvRV(sv);
238b27b3
NC
1249 old_type_details = &fake_rv;
1250 if (new_type == SVt_NV)
1251 new_type = SVt_PVNV;
4df7f6af
NC
1252 } else {
1253 if (new_type < SVt_PVIV) {
1254 new_type = (new_type == SVt_NV)
1255 ? SVt_PVNV : SVt_PVIV;
1256 }
bd81e77b
NC
1257 }
1258 break;
1259 case SVt_NV:
1260 if (new_type < SVt_PVNV) {
1261 new_type = SVt_PVNV;
bd81e77b
NC
1262 }
1263 break;
bd81e77b
NC
1264 case SVt_PV:
1265 assert(new_type > SVt_PV);
1266 assert(SVt_IV < SVt_PV);
1267 assert(SVt_NV < SVt_PV);
1268 break;
1269 case SVt_PVIV:
1270 break;
1271 case SVt_PVNV:
1272 break;
1273 case SVt_PVMG:
1274 /* Because the XPVMG of PL_mess_sv isn't allocated from the arena,
1275 there's no way that it can be safely upgraded, because perl.c
1276 expects to Safefree(SvANY(PL_mess_sv)) */
1277 assert(sv != PL_mess_sv);
1278 /* This flag bit is used to mean other things in other scalar types.
1279 Given that it only has meaning inside the pad, it shouldn't be set
1280 on anything that can get upgraded. */
00b1698f 1281 assert(!SvPAD_TYPED(sv));
bd81e77b
NC
1282 break;
1283 default:
1284 if (old_type_details->cant_upgrade)
c81225bc
NC
1285 Perl_croak(aTHX_ "Can't upgrade %s (%" UVuf ") to %" UVuf,
1286 sv_reftype(sv, 0), (UV) old_type, (UV) new_type);
bd81e77b 1287 }
3376de98
NC
1288
1289 if (old_type > new_type)
1290 Perl_croak(aTHX_ "sv_upgrade from type %d down to type %d",
1291 (int)old_type, (int)new_type);
1292
2fa1109b 1293 new_type_details = bodies_by_type + new_type;
645c22ef 1294
bd81e77b
NC
1295 SvFLAGS(sv) &= ~SVTYPEMASK;
1296 SvFLAGS(sv) |= new_type;
932e9ff9 1297
ab4416c0
NC
1298 /* This can't happen, as SVt_NULL is <= all values of new_type, so one of
1299 the return statements above will have triggered. */
1300 assert (new_type != SVt_NULL);
bd81e77b 1301 switch (new_type) {
bd81e77b
NC
1302 case SVt_IV:
1303 assert(old_type == SVt_NULL);
1304 SvANY(sv) = (XPVIV*)((char*)&(sv->sv_u.svu_iv) - STRUCT_OFFSET(XPVIV, xiv_iv));
1305 SvIV_set(sv, 0);
1306 return;
1307 case SVt_NV:
1308 assert(old_type == SVt_NULL);
1309 SvANY(sv) = new_XNV();
1310 SvNV_set(sv, 0);
1311 return;
bd81e77b 1312 case SVt_PVHV:
bd81e77b 1313 case SVt_PVAV:
d2a0f284 1314 assert(new_type_details->body_size);
c1ae03ae
NC
1315
1316#ifndef PURIFY
1317 assert(new_type_details->arena);
d2a0f284 1318 assert(new_type_details->arena_size);
c1ae03ae 1319 /* This points to the start of the allocated area. */
d2a0f284
JC
1320 new_body_inline(new_body, new_type);
1321 Zero(new_body, new_type_details->body_size, char);
c1ae03ae
NC
1322 new_body = ((char *)new_body) - new_type_details->offset;
1323#else
1324 /* We always allocated the full length item with PURIFY. To do this
1325 we fake things so that arena is false for all 16 types.. */
1326 new_body = new_NOARENAZ(new_type_details);
1327#endif
1328 SvANY(sv) = new_body;
1329 if (new_type == SVt_PVAV) {
1330 AvMAX(sv) = -1;
1331 AvFILLp(sv) = -1;
1332 AvREAL_only(sv);
64484faa 1333 if (old_type_details->body_size) {
ac572bf4
NC
1334 AvALLOC(sv) = 0;
1335 } else {
1336 /* It will have been zeroed when the new body was allocated.
1337 Lets not write to it, in case it confuses a write-back
1338 cache. */
1339 }
78ac7dd9
NC
1340 } else {
1341 assert(!SvOK(sv));
1342 SvOK_off(sv);
1343#ifndef NODEFAULT_SHAREKEYS
1344 HvSHAREKEYS_on(sv); /* key-sharing on by default */
1345#endif
1346 HvMAX(sv) = 7; /* (start with 8 buckets) */
64484faa 1347 if (old_type_details->body_size) {
78ac7dd9
NC
1348 HvFILL(sv) = 0;
1349 } else {
1350 /* It will have been zeroed when the new body was allocated.
1351 Lets not write to it, in case it confuses a write-back
1352 cache. */
1353 }
c1ae03ae 1354 }
aeb18a1e 1355
bd81e77b
NC
1356 /* SVt_NULL isn't the only thing upgraded to AV or HV.
1357 The target created by newSVrv also is, and it can have magic.
1358 However, it never has SvPVX set.
1359 */
4df7f6af
NC
1360 if (old_type == SVt_IV) {
1361 assert(!SvROK(sv));
1362 } else if (old_type >= SVt_PV) {
bd81e77b
NC
1363 assert(SvPVX_const(sv) == 0);
1364 }
aeb18a1e 1365
bd81e77b 1366 if (old_type >= SVt_PVMG) {
e736a858 1367 SvMAGIC_set(sv, ((XPVMG*)old_body)->xmg_u.xmg_magic);
bd81e77b 1368 SvSTASH_set(sv, ((XPVMG*)old_body)->xmg_stash);
797c7171
NC
1369 } else {
1370 sv->sv_u.svu_array = NULL; /* or svu_hash */
bd81e77b
NC
1371 }
1372 break;
93e68bfb 1373
93e68bfb 1374
bd81e77b
NC
1375 case SVt_PVIV:
1376 /* XXX Is this still needed? Was it ever needed? Surely as there is
1377 no route from NV to PVIV, NOK can never be true */
1378 assert(!SvNOKp(sv));
1379 assert(!SvNOK(sv));
1380 case SVt_PVIO:
1381 case SVt_PVFM:
bd81e77b
NC
1382 case SVt_PVGV:
1383 case SVt_PVCV:
1384 case SVt_PVLV:
5c35adbb 1385 case SVt_REGEXP:
bd81e77b
NC
1386 case SVt_PVMG:
1387 case SVt_PVNV:
1388 case SVt_PV:
93e68bfb 1389
d2a0f284 1390 assert(new_type_details->body_size);
bd81e77b
NC
1391 /* We always allocated the full length item with PURIFY. To do this
1392 we fake things so that arena is false for all 16 types.. */
1393 if(new_type_details->arena) {
1394 /* This points to the start of the allocated area. */
d2a0f284
JC
1395 new_body_inline(new_body, new_type);
1396 Zero(new_body, new_type_details->body_size, char);
bd81e77b
NC
1397 new_body = ((char *)new_body) - new_type_details->offset;
1398 } else {
1399 new_body = new_NOARENAZ(new_type_details);
1400 }
1401 SvANY(sv) = new_body;
5e2fc214 1402
bd81e77b 1403 if (old_type_details->copy) {
f9ba3d20
NC
1404 /* There is now the potential for an upgrade from something without
1405 an offset (PVNV or PVMG) to something with one (PVCV, PVFM) */
1406 int offset = old_type_details->offset;
1407 int length = old_type_details->copy;
1408
1409 if (new_type_details->offset > old_type_details->offset) {
d4c19fe8 1410 const int difference
f9ba3d20
NC
1411 = new_type_details->offset - old_type_details->offset;
1412 offset += difference;
1413 length -= difference;
1414 }
1415 assert (length >= 0);
1416
1417 Copy((char *)old_body + offset, (char *)new_body + offset, length,
1418 char);
bd81e77b
NC
1419 }
1420
1421#ifndef NV_ZERO_IS_ALLBITS_ZERO
f2524eef 1422 /* If NV 0.0 is stores as all bits 0 then Zero() already creates a
e5ce394c
NC
1423 * correct 0.0 for us. Otherwise, if the old body didn't have an
1424 * NV slot, but the new one does, then we need to initialise the
1425 * freshly created NV slot with whatever the correct bit pattern is
1426 * for 0.0 */
e22a937e
NC
1427 if (old_type_details->zero_nv && !new_type_details->zero_nv
1428 && !isGV_with_GP(sv))
bd81e77b 1429 SvNV_set(sv, 0);
82048762 1430#endif
5e2fc214 1431
85dca89a
NC
1432 if (new_type == SVt_PVIO) {
1433 IO * const io = MUTABLE_IO(sv);
31c9a3ac 1434 GV *iogv = gv_fetchpvs("IO::Handle::", GV_ADD, SVt_PVHV);
85dca89a
NC
1435
1436 SvOBJECT_on(io);
1437 /* Clear the stashcache because a new IO could overrule a package
1438 name */
1439 hv_clear(PL_stashcache);
1440
85dca89a 1441 SvSTASH_set(io, MUTABLE_HV(SvREFCNT_inc(GvHV(iogv))));
f2524eef 1442 IoPAGE_LEN(sv) = 60;
85dca89a 1443 }
4df7f6af
NC
1444 if (old_type < SVt_PV) {
1445 /* referant will be NULL unless the old type was SVt_IV emulating
1446 SVt_RV */
1447 sv->sv_u.svu_rv = referant;
1448 }
bd81e77b
NC
1449 break;
1450 default:
afd78fd5
JH
1451 Perl_croak(aTHX_ "panic: sv_upgrade to unknown type %lu",
1452 (unsigned long)new_type);
bd81e77b 1453 }
73171d91 1454
bc786448 1455 if (old_type > SVt_IV) { /* SVt_IVs are overloaded for PTEs */
bd81e77b
NC
1456#ifdef PURIFY
1457 my_safefree(old_body);
1458#else
bc786448
GG
1459 /* Note that there is an assumption that all bodies of types that
1460 can be upgraded came from arenas. Only the more complex non-
1461 upgradable types are allowed to be directly malloc()ed. */
1462 assert(old_type_details->arena);
bd81e77b
NC
1463 del_body((void*)((char*)old_body + old_type_details->offset),
1464 &PL_body_roots[old_type]);
1465#endif
1466 }
1467}
73171d91 1468
bd81e77b
NC
1469/*
1470=for apidoc sv_backoff
73171d91 1471
bd81e77b
NC
1472Remove any string offset. You should normally use the C<SvOOK_off> macro
1473wrapper instead.
73171d91 1474
bd81e77b 1475=cut
73171d91
NC
1476*/
1477
bd81e77b 1478int
aad570aa 1479Perl_sv_backoff(pTHX_ register SV *const sv)
bd81e77b 1480{
69240efd 1481 STRLEN delta;
7a4bba22 1482 const char * const s = SvPVX_const(sv);
7918f24d
NC
1483
1484 PERL_ARGS_ASSERT_SV_BACKOFF;
96a5add6 1485 PERL_UNUSED_CONTEXT;
7918f24d 1486
bd81e77b
NC
1487 assert(SvOOK(sv));
1488 assert(SvTYPE(sv) != SVt_PVHV);
1489 assert(SvTYPE(sv) != SVt_PVAV);
7a4bba22 1490
69240efd
NC
1491 SvOOK_offset(sv, delta);
1492
7a4bba22
NC
1493 SvLEN_set(sv, SvLEN(sv) + delta);
1494 SvPV_set(sv, SvPVX(sv) - delta);
1495 Move(s, SvPVX(sv), SvCUR(sv)+1, char);
bd81e77b
NC
1496 SvFLAGS(sv) &= ~SVf_OOK;
1497 return 0;
1498}
73171d91 1499
bd81e77b
NC
1500/*
1501=for apidoc sv_grow
73171d91 1502
bd81e77b
NC
1503Expands the character buffer in the SV. If necessary, uses C<sv_unref> and
1504upgrades the SV to C<SVt_PV>. Returns a pointer to the character buffer.
1505Use the C<SvGROW> wrapper instead.
93e68bfb 1506
bd81e77b
NC
1507=cut
1508*/
93e68bfb 1509
bd81e77b 1510char *
aad570aa 1511Perl_sv_grow(pTHX_ register SV *const sv, register STRLEN newlen)
bd81e77b
NC
1512{
1513 register char *s;
93e68bfb 1514
7918f24d
NC
1515 PERL_ARGS_ASSERT_SV_GROW;
1516
5db06880
NC
1517 if (PL_madskills && newlen >= 0x100000) {
1518 PerlIO_printf(Perl_debug_log,
1519 "Allocation too large: %"UVxf"\n", (UV)newlen);
1520 }
bd81e77b
NC
1521#ifdef HAS_64K_LIMIT
1522 if (newlen >= 0x10000) {
1523 PerlIO_printf(Perl_debug_log,
1524 "Allocation too large: %"UVxf"\n", (UV)newlen);
1525 my_exit(1);
1526 }
1527#endif /* HAS_64K_LIMIT */
1528 if (SvROK(sv))
1529 sv_unref(sv);
1530 if (SvTYPE(sv) < SVt_PV) {
1531 sv_upgrade(sv, SVt_PV);
1532 s = SvPVX_mutable(sv);
1533 }
1534 else if (SvOOK(sv)) { /* pv is offset? */
1535 sv_backoff(sv);
1536 s = SvPVX_mutable(sv);
1537 if (newlen > SvLEN(sv))
1538 newlen += 10 * (newlen - SvCUR(sv)); /* avoid copy each time */
1539#ifdef HAS_64K_LIMIT
1540 if (newlen >= 0x10000)
1541 newlen = 0xFFFF;
1542#endif
1543 }
1544 else
1545 s = SvPVX_mutable(sv);
aeb18a1e 1546
bd81e77b 1547 if (newlen > SvLEN(sv)) { /* need more room? */
aedff202 1548#ifndef Perl_safesysmalloc_size
bd81e77b 1549 newlen = PERL_STRLEN_ROUNDUP(newlen);
bd81e77b 1550#endif
98653f18 1551 if (SvLEN(sv) && s) {
10edeb5d 1552 s = (char*)saferealloc(s, newlen);
bd81e77b
NC
1553 }
1554 else {
10edeb5d 1555 s = (char*)safemalloc(newlen);
bd81e77b
NC
1556 if (SvPVX_const(sv) && SvCUR(sv)) {
1557 Move(SvPVX_const(sv), s, (newlen < SvCUR(sv)) ? newlen : SvCUR(sv), char);
1558 }
1559 }
1560 SvPV_set(sv, s);
ca7c1a29 1561#ifdef Perl_safesysmalloc_size
98653f18
NC
1562 /* Do this here, do it once, do it right, and then we will never get
1563 called back into sv_grow() unless there really is some growing
1564 needed. */
ca7c1a29 1565 SvLEN_set(sv, Perl_safesysmalloc_size(s));
98653f18 1566#else
bd81e77b 1567 SvLEN_set(sv, newlen);
98653f18 1568#endif
bd81e77b
NC
1569 }
1570 return s;
1571}
aeb18a1e 1572
bd81e77b
NC
1573/*
1574=for apidoc sv_setiv
932e9ff9 1575
bd81e77b
NC
1576Copies an integer into the given SV, upgrading first if necessary.
1577Does not handle 'set' magic. See also C<sv_setiv_mg>.
463ee0b2 1578
bd81e77b
NC
1579=cut
1580*/
463ee0b2 1581
bd81e77b 1582void
aad570aa 1583Perl_sv_setiv(pTHX_ register SV *const sv, const IV i)
bd81e77b 1584{
97aff369 1585 dVAR;
7918f24d
NC
1586
1587 PERL_ARGS_ASSERT_SV_SETIV;
1588
bd81e77b
NC
1589 SV_CHECK_THINKFIRST_COW_DROP(sv);
1590 switch (SvTYPE(sv)) {
1591 case SVt_NULL:
bd81e77b 1592 case SVt_NV:
3376de98 1593 sv_upgrade(sv, SVt_IV);
bd81e77b 1594 break;
bd81e77b
NC
1595 case SVt_PV:
1596 sv_upgrade(sv, SVt_PVIV);
1597 break;
463ee0b2 1598
bd81e77b 1599 case SVt_PVGV:
6e592b3a
BM
1600 if (!isGV_with_GP(sv))
1601 break;
bd81e77b
NC
1602 case SVt_PVAV:
1603 case SVt_PVHV:
1604 case SVt_PVCV:
1605 case SVt_PVFM:
1606 case SVt_PVIO:
1607 Perl_croak(aTHX_ "Can't coerce %s to integer in %s", sv_reftype(sv,0),
1608 OP_DESC(PL_op));
42d0e0b7 1609 default: NOOP;
bd81e77b
NC
1610 }
1611 (void)SvIOK_only(sv); /* validate number */
1612 SvIV_set(sv, i);
1613 SvTAINT(sv);
1614}
932e9ff9 1615
bd81e77b
NC
1616/*
1617=for apidoc sv_setiv_mg
d33b2eba 1618
bd81e77b 1619Like C<sv_setiv>, but also handles 'set' magic.
1c846c1f 1620
bd81e77b
NC
1621=cut
1622*/
d33b2eba 1623
bd81e77b 1624void
aad570aa 1625Perl_sv_setiv_mg(pTHX_ register SV *const sv, const IV i)
bd81e77b 1626{
7918f24d
NC
1627 PERL_ARGS_ASSERT_SV_SETIV_MG;
1628
bd81e77b
NC
1629 sv_setiv(sv,i);
1630 SvSETMAGIC(sv);
1631}
727879eb 1632
bd81e77b
NC
1633/*
1634=for apidoc sv_setuv
d33b2eba 1635
bd81e77b
NC
1636Copies an unsigned integer into the given SV, upgrading first if necessary.
1637Does not handle 'set' magic. See also C<sv_setuv_mg>.
9b94d1dd 1638
bd81e77b
NC
1639=cut
1640*/
d33b2eba 1641
bd81e77b 1642void
aad570aa 1643Perl_sv_setuv(pTHX_ register SV *const sv, const UV u)
bd81e77b 1644{
7918f24d
NC
1645 PERL_ARGS_ASSERT_SV_SETUV;
1646
bd81e77b
NC
1647 /* With these two if statements:
1648 u=1.49 s=0.52 cu=72.49 cs=10.64 scripts=270 tests=20865
d33b2eba 1649
bd81e77b
NC
1650 without
1651 u=1.35 s=0.47 cu=73.45 cs=11.43 scripts=270 tests=20865
1c846c1f 1652
bd81e77b
NC
1653 If you wish to remove them, please benchmark to see what the effect is
1654 */
1655 if (u <= (UV)IV_MAX) {
1656 sv_setiv(sv, (IV)u);
1657 return;
1658 }
1659 sv_setiv(sv, 0);
1660 SvIsUV_on(sv);
1661 SvUV_set(sv, u);
1662}
d33b2eba 1663
bd81e77b
NC
1664/*
1665=for apidoc sv_setuv_mg
727879eb 1666
bd81e77b 1667Like C<sv_setuv>, but also handles 'set' magic.
9b94d1dd 1668
bd81e77b
NC
1669=cut
1670*/
5e2fc214 1671
bd81e77b 1672void
aad570aa 1673Perl_sv_setuv_mg(pTHX_ register SV *const sv, const UV u)
bd81e77b 1674{
7918f24d
NC
1675 PERL_ARGS_ASSERT_SV_SETUV_MG;
1676
bd81e77b
NC
1677 sv_setuv(sv,u);
1678 SvSETMAGIC(sv);
1679}
5e2fc214 1680
954c1994 1681/*
bd81e77b 1682=for apidoc sv_setnv
954c1994 1683
bd81e77b
NC
1684Copies a double into the given SV, upgrading first if necessary.
1685Does not handle 'set' magic. See also C<sv_setnv_mg>.
954c1994
GS
1686
1687=cut
1688*/
1689
63f97190 1690void
aad570aa 1691Perl_sv_setnv(pTHX_ register SV *const sv, const NV num)
79072805 1692{
97aff369 1693 dVAR;
7918f24d
NC
1694
1695 PERL_ARGS_ASSERT_SV_SETNV;
1696
bd81e77b
NC
1697 SV_CHECK_THINKFIRST_COW_DROP(sv);
1698 switch (SvTYPE(sv)) {
79072805 1699 case SVt_NULL:
79072805 1700 case SVt_IV:
bd81e77b 1701 sv_upgrade(sv, SVt_NV);
79072805
LW
1702 break;
1703 case SVt_PV:
79072805 1704 case SVt_PVIV:
bd81e77b 1705 sv_upgrade(sv, SVt_PVNV);
79072805 1706 break;
bd4b1eb5 1707
bd4b1eb5 1708 case SVt_PVGV:
6e592b3a
BM
1709 if (!isGV_with_GP(sv))
1710 break;
bd81e77b
NC
1711 case SVt_PVAV:
1712 case SVt_PVHV:
79072805 1713 case SVt_PVCV:
bd81e77b
NC
1714 case SVt_PVFM:
1715 case SVt_PVIO:
1716 Perl_croak(aTHX_ "Can't coerce %s to number in %s", sv_reftype(sv,0),
1717 OP_NAME(PL_op));
42d0e0b7 1718 default: NOOP;
2068cd4d 1719 }
bd81e77b
NC
1720 SvNV_set(sv, num);
1721 (void)SvNOK_only(sv); /* validate number */
1722 SvTAINT(sv);
79072805
LW
1723}
1724
645c22ef 1725/*
bd81e77b 1726=for apidoc sv_setnv_mg
645c22ef 1727
bd81e77b 1728Like C<sv_setnv>, but also handles 'set' magic.
645c22ef
DM
1729
1730=cut
1731*/
1732
bd81e77b 1733void
aad570aa 1734Perl_sv_setnv_mg(pTHX_ register SV *const sv, const NV num)
79072805 1735{
7918f24d
NC
1736 PERL_ARGS_ASSERT_SV_SETNV_MG;
1737
bd81e77b
NC
1738 sv_setnv(sv,num);
1739 SvSETMAGIC(sv);
79072805
LW
1740}
1741
bd81e77b
NC
1742/* Print an "isn't numeric" warning, using a cleaned-up,
1743 * printable version of the offending string
1744 */
954c1994 1745
bd81e77b 1746STATIC void
aad570aa 1747S_not_a_number(pTHX_ SV *const sv)
79072805 1748{
97aff369 1749 dVAR;
bd81e77b
NC
1750 SV *dsv;
1751 char tmpbuf[64];
1752 const char *pv;
94463019 1753
7918f24d
NC
1754 PERL_ARGS_ASSERT_NOT_A_NUMBER;
1755
94463019 1756 if (DO_UTF8(sv)) {
84bafc02 1757 dsv = newSVpvs_flags("", SVs_TEMP);
94463019
JH
1758 pv = sv_uni_display(dsv, sv, 10, 0);
1759 } else {
1760 char *d = tmpbuf;
551405c4 1761 const char * const limit = tmpbuf + sizeof(tmpbuf) - 8;
94463019
JH
1762 /* each *s can expand to 4 chars + "...\0",
1763 i.e. need room for 8 chars */
ecdeb87c 1764
00b6aa41
AL
1765 const char *s = SvPVX_const(sv);
1766 const char * const end = s + SvCUR(sv);
1767 for ( ; s < end && d < limit; s++ ) {
94463019
JH
1768 int ch = *s & 0xFF;
1769 if (ch & 128 && !isPRINT_LC(ch)) {
1770 *d++ = 'M';
1771 *d++ = '-';
1772 ch &= 127;
1773 }
1774 if (ch == '\n') {
1775 *d++ = '\\';
1776 *d++ = 'n';
1777 }
1778 else if (ch == '\r') {
1779 *d++ = '\\';
1780 *d++ = 'r';
1781 }
1782 else if (ch == '\f') {
1783 *d++ = '\\';
1784 *d++ = 'f';
1785 }
1786 else if (ch == '\\') {
1787 *d++ = '\\';
1788 *d++ = '\\';
1789 }
1790 else if (ch == '\0') {
1791 *d++ = '\\';
1792 *d++ = '0';
1793 }
1794 else if (isPRINT_LC(ch))
1795 *d++ = ch;
1796 else {
1797 *d++ = '^';
1798 *d++ = toCTRL(ch);
1799 }
1800 }
1801 if (s < end) {
1802 *d++ = '.';
1803 *d++ = '.';
1804 *d++ = '.';
1805 }
1806 *d = '\0';
1807 pv = tmpbuf;
a0d0e21e 1808 }
a0d0e21e 1809
533c011a 1810 if (PL_op)
9014280d 1811 Perl_warner(aTHX_ packWARN(WARN_NUMERIC),
94463019
JH
1812 "Argument \"%s\" isn't numeric in %s", pv,
1813 OP_DESC(PL_op));
a0d0e21e 1814 else
9014280d 1815 Perl_warner(aTHX_ packWARN(WARN_NUMERIC),
94463019 1816 "Argument \"%s\" isn't numeric", pv);
a0d0e21e
LW
1817}
1818
c2988b20
NC
1819/*
1820=for apidoc looks_like_number
1821
645c22ef
DM
1822Test if the content of an SV looks like a number (or is a number).
1823C<Inf> and C<Infinity> are treated as numbers (so will not issue a
1824non-numeric warning), even if your atof() doesn't grok them.
c2988b20
NC
1825
1826=cut
1827*/
1828
1829I32
aad570aa 1830Perl_looks_like_number(pTHX_ SV *const sv)
c2988b20 1831{
a3b680e6 1832 register const char *sbegin;
c2988b20
NC
1833 STRLEN len;
1834
7918f24d
NC
1835 PERL_ARGS_ASSERT_LOOKS_LIKE_NUMBER;
1836
c2988b20 1837 if (SvPOK(sv)) {
3f7c398e 1838 sbegin = SvPVX_const(sv);
c2988b20
NC
1839 len = SvCUR(sv);
1840 }
1841 else if (SvPOKp(sv))
83003860 1842 sbegin = SvPV_const(sv, len);
c2988b20 1843 else
e0ab1c0e 1844 return SvFLAGS(sv) & (SVf_NOK|SVp_NOK|SVf_IOK|SVp_IOK);
c2988b20
NC
1845 return grok_number(sbegin, len, NULL);
1846}
25da4f38 1847
19f6321d
NC
1848STATIC bool
1849S_glob_2number(pTHX_ GV * const gv)
180488f8
NC
1850{
1851 const U32 wasfake = SvFLAGS(gv) & SVf_FAKE;
1852 SV *const buffer = sv_newmortal();
1853
7918f24d
NC
1854 PERL_ARGS_ASSERT_GLOB_2NUMBER;
1855
180488f8
NC
1856 /* FAKE globs can get coerced, so need to turn this off temporarily if it
1857 is on. */
1858 SvFAKE_off(gv);
1859 gv_efullname3(buffer, gv, "*");
1860 SvFLAGS(gv) |= wasfake;
1861
675c862f
AL
1862 /* We know that all GVs stringify to something that is not-a-number,
1863 so no need to test that. */
1864 if (ckWARN(WARN_NUMERIC))
1865 not_a_number(buffer);
1866 /* We just want something true to return, so that S_sv_2iuv_common
1867 can tail call us and return true. */
19f6321d 1868 return TRUE;
675c862f
AL
1869}
1870
25da4f38
IZ
1871/* Actually, ISO C leaves conversion of UV to IV undefined, but
1872 until proven guilty, assume that things are not that bad... */
1873
645c22ef
DM
1874/*
1875 NV_PRESERVES_UV:
1876
1877 As 64 bit platforms often have an NV that doesn't preserve all bits of
28e5dec8
JH
1878 an IV (an assumption perl has been based on to date) it becomes necessary
1879 to remove the assumption that the NV always carries enough precision to
1880 recreate the IV whenever needed, and that the NV is the canonical form.
1881 Instead, IV/UV and NV need to be given equal rights. So as to not lose
645c22ef 1882 precision as a side effect of conversion (which would lead to insanity
28e5dec8
JH
1883 and the dragon(s) in t/op/numconvert.t getting very angry) the intent is
1884 1) to distinguish between IV/UV/NV slots that have cached a valid
1885 conversion where precision was lost and IV/UV/NV slots that have a
1886 valid conversion which has lost no precision
645c22ef 1887 2) to ensure that if a numeric conversion to one form is requested that
28e5dec8
JH
1888 would lose precision, the precise conversion (or differently
1889 imprecise conversion) is also performed and cached, to prevent
1890 requests for different numeric formats on the same SV causing
1891 lossy conversion chains. (lossless conversion chains are perfectly
1892 acceptable (still))
1893
1894
1895 flags are used:
1896 SvIOKp is true if the IV slot contains a valid value
1897 SvIOK is true only if the IV value is accurate (UV if SvIOK_UV true)
1898 SvNOKp is true if the NV slot contains a valid value
1899 SvNOK is true only if the NV value is accurate
1900
1901 so
645c22ef 1902 while converting from PV to NV, check to see if converting that NV to an
28e5dec8
JH
1903 IV(or UV) would lose accuracy over a direct conversion from PV to
1904 IV(or UV). If it would, cache both conversions, return NV, but mark
1905 SV as IOK NOKp (ie not NOK).
1906
645c22ef 1907 While converting from PV to IV, check to see if converting that IV to an
28e5dec8
JH
1908 NV would lose accuracy over a direct conversion from PV to NV. If it
1909 would, cache both conversions, flag similarly.
1910
1911 Before, the SV value "3.2" could become NV=3.2 IV=3 NOK, IOK quite
1912 correctly because if IV & NV were set NV *always* overruled.
645c22ef
DM
1913 Now, "3.2" will become NV=3.2 IV=3 NOK, IOKp, because the flag's meaning
1914 changes - now IV and NV together means that the two are interchangeable:
28e5dec8 1915 SvIVX == (IV) SvNVX && SvNVX == (NV) SvIVX;
d460ef45 1916
645c22ef
DM
1917 The benefit of this is that operations such as pp_add know that if
1918 SvIOK is true for both left and right operands, then integer addition
1919 can be used instead of floating point (for cases where the result won't
1920 overflow). Before, floating point was always used, which could lead to
28e5dec8
JH
1921 loss of precision compared with integer addition.
1922
1923 * making IV and NV equal status should make maths accurate on 64 bit
1924 platforms
1925 * may speed up maths somewhat if pp_add and friends start to use
645c22ef 1926 integers when possible instead of fp. (Hopefully the overhead in
28e5dec8
JH
1927 looking for SvIOK and checking for overflow will not outweigh the
1928 fp to integer speedup)
1929 * will slow down integer operations (callers of SvIV) on "inaccurate"
1930 values, as the change from SvIOK to SvIOKp will cause a call into
1931 sv_2iv each time rather than a macro access direct to the IV slot
1932 * should speed up number->string conversion on integers as IV is
645c22ef 1933 favoured when IV and NV are equally accurate
28e5dec8
JH
1934
1935 ####################################################################
645c22ef
DM
1936 You had better be using SvIOK_notUV if you want an IV for arithmetic:
1937 SvIOK is true if (IV or UV), so you might be getting (IV)SvUV.
1938 On the other hand, SvUOK is true iff UV.
28e5dec8
JH
1939 ####################################################################
1940
645c22ef 1941 Your mileage will vary depending your CPU's relative fp to integer
28e5dec8
JH
1942 performance ratio.
1943*/
1944
1945#ifndef NV_PRESERVES_UV
645c22ef
DM
1946# define IS_NUMBER_UNDERFLOW_IV 1
1947# define IS_NUMBER_UNDERFLOW_UV 2
1948# define IS_NUMBER_IV_AND_UV 2
1949# define IS_NUMBER_OVERFLOW_IV 4
1950# define IS_NUMBER_OVERFLOW_UV 5
1951
1952/* sv_2iuv_non_preserve(): private routine for use by sv_2iv() and sv_2uv() */
28e5dec8
JH
1953
1954/* For sv_2nv these three cases are "SvNOK and don't bother casting" */
1955STATIC int
5de3775c 1956S_sv_2iuv_non_preserve(pTHX_ register SV *const sv
47031da6
NC
1957# ifdef DEBUGGING
1958 , I32 numtype
1959# endif
1960 )
28e5dec8 1961{
97aff369 1962 dVAR;
7918f24d
NC
1963
1964 PERL_ARGS_ASSERT_SV_2IUV_NON_PRESERVE;
1965
3f7c398e 1966 DEBUG_c(PerlIO_printf(Perl_debug_log,"sv_2iuv_non '%s', IV=0x%"UVxf" NV=%"NVgf" inttype=%"UVXf"\n", SvPVX_const(sv), SvIVX(sv), SvNVX(sv), (UV)numtype));
28e5dec8
JH
1967 if (SvNVX(sv) < (NV)IV_MIN) {
1968 (void)SvIOKp_on(sv);
1969 (void)SvNOK_on(sv);
45977657 1970 SvIV_set(sv, IV_MIN);
28e5dec8
JH
1971 return IS_NUMBER_UNDERFLOW_IV;
1972 }
1973 if (SvNVX(sv) > (NV)UV_MAX) {
1974 (void)SvIOKp_on(sv);
1975 (void)SvNOK_on(sv);
1976 SvIsUV_on(sv);
607fa7f2 1977 SvUV_set(sv, UV_MAX);
28e5dec8
JH
1978 return IS_NUMBER_OVERFLOW_UV;
1979 }
c2988b20
NC
1980 (void)SvIOKp_on(sv);
1981 (void)SvNOK_on(sv);
1982 /* Can't use strtol etc to convert this string. (See truth table in
1983 sv_2iv */
1984 if (SvNVX(sv) <= (UV)IV_MAX) {
45977657 1985 SvIV_set(sv, I_V(SvNVX(sv)));
c2988b20
NC
1986 if ((NV)(SvIVX(sv)) == SvNVX(sv)) {
1987 SvIOK_on(sv); /* Integer is precise. NOK, IOK */
1988 } else {
1989 /* Integer is imprecise. NOK, IOKp */
1990 }
1991 return SvNVX(sv) < 0 ? IS_NUMBER_UNDERFLOW_UV : IS_NUMBER_IV_AND_UV;
1992 }
1993 SvIsUV_on(sv);
607fa7f2 1994 SvUV_set(sv, U_V(SvNVX(sv)));
c2988b20
NC
1995 if ((NV)(SvUVX(sv)) == SvNVX(sv)) {
1996 if (SvUVX(sv) == UV_MAX) {
1997 /* As we know that NVs don't preserve UVs, UV_MAX cannot
1998 possibly be preserved by NV. Hence, it must be overflow.
1999 NOK, IOKp */
2000 return IS_NUMBER_OVERFLOW_UV;
2001 }
2002 SvIOK_on(sv); /* Integer is precise. NOK, UOK */
2003 } else {
2004 /* Integer is imprecise. NOK, IOKp */
28e5dec8 2005 }
c2988b20 2006 return IS_NUMBER_OVERFLOW_IV;
28e5dec8 2007}
645c22ef
DM
2008#endif /* !NV_PRESERVES_UV*/
2009
af359546 2010STATIC bool
7918f24d
NC
2011S_sv_2iuv_common(pTHX_ SV *const sv)
2012{
97aff369 2013 dVAR;
7918f24d
NC
2014
2015 PERL_ARGS_ASSERT_SV_2IUV_COMMON;
2016
af359546 2017 if (SvNOKp(sv)) {
28e5dec8
JH
2018 /* erm. not sure. *should* never get NOKp (without NOK) from sv_2nv
2019 * without also getting a cached IV/UV from it at the same time
2020 * (ie PV->NV conversion should detect loss of accuracy and cache
af359546
NC
2021 * IV or UV at same time to avoid this. */
2022 /* IV-over-UV optimisation - choose to cache IV if possible */
25da4f38
IZ
2023
2024 if (SvTYPE(sv) == SVt_NV)
2025 sv_upgrade(sv, SVt_PVNV);
2026
28e5dec8
JH
2027 (void)SvIOKp_on(sv); /* Must do this first, to clear any SvOOK */
2028 /* < not <= as for NV doesn't preserve UV, ((NV)IV_MAX+1) will almost
2029 certainly cast into the IV range at IV_MAX, whereas the correct
2030 answer is the UV IV_MAX +1. Hence < ensures that dodgy boundary
2031 cases go to UV */
cab190d4
JD
2032#if defined(NAN_COMPARE_BROKEN) && defined(Perl_isnan)
2033 if (Perl_isnan(SvNVX(sv))) {
2034 SvUV_set(sv, 0);
2035 SvIsUV_on(sv);
fdbe6d7c 2036 return FALSE;
cab190d4 2037 }
cab190d4 2038#endif
28e5dec8 2039 if (SvNVX(sv) < (NV)IV_MAX + 0.5) {
45977657 2040 SvIV_set(sv, I_V(SvNVX(sv)));
28e5dec8
JH
2041 if (SvNVX(sv) == (NV) SvIVX(sv)
2042#ifndef NV_PRESERVES_UV
2043 && (((UV)1 << NV_PRESERVES_UV_BITS) >
2044 (UV)(SvIVX(sv) > 0 ? SvIVX(sv) : -SvIVX(sv)))
2045 /* Don't flag it as "accurately an integer" if the number
2046 came from a (by definition imprecise) NV operation, and
2047 we're outside the range of NV integer precision */
2048#endif
2049 ) {
a43d94f2
NC
2050 if (SvNOK(sv))
2051 SvIOK_on(sv); /* Can this go wrong with rounding? NWC */
2052 else {
2053 /* scalar has trailing garbage, eg "42a" */
2054 }
28e5dec8 2055 DEBUG_c(PerlIO_printf(Perl_debug_log,
7234c960 2056 "0x%"UVxf" iv(%"NVgf" => %"IVdf") (precise)\n",
28e5dec8
JH
2057 PTR2UV(sv),
2058 SvNVX(sv),
2059 SvIVX(sv)));
2060
2061 } else {
2062 /* IV not precise. No need to convert from PV, as NV
2063 conversion would already have cached IV if it detected
2064 that PV->IV would be better than PV->NV->IV
2065 flags already correct - don't set public IOK. */
2066 DEBUG_c(PerlIO_printf(Perl_debug_log,
7234c960 2067 "0x%"UVxf" iv(%"NVgf" => %"IVdf") (imprecise)\n",
28e5dec8
JH
2068 PTR2UV(sv),
2069 SvNVX(sv),
2070 SvIVX(sv)));
2071 }
2072 /* Can the above go wrong if SvIVX == IV_MIN and SvNVX < IV_MIN,
2073 but the cast (NV)IV_MIN rounds to a the value less (more
2074 negative) than IV_MIN which happens to be equal to SvNVX ??
2075 Analogous to 0xFFFFFFFFFFFFFFFF rounding up to NV (2**64) and
2076 NV rounding back to 0xFFFFFFFFFFFFFFFF, so UVX == UV(NVX) and
2077 (NV)UVX == NVX are both true, but the values differ. :-(
2078 Hopefully for 2s complement IV_MIN is something like
2079 0x8000000000000000 which will be exact. NWC */
d460ef45 2080 }
25da4f38 2081 else {
607fa7f2 2082 SvUV_set(sv, U_V(SvNVX(sv)));
28e5dec8
JH
2083 if (
2084 (SvNVX(sv) == (NV) SvUVX(sv))
2085#ifndef NV_PRESERVES_UV
2086 /* Make sure it's not 0xFFFFFFFFFFFFFFFF */
2087 /*&& (SvUVX(sv) != UV_MAX) irrelevant with code below */
2088 && (((UV)1 << NV_PRESERVES_UV_BITS) > SvUVX(sv))
2089 /* Don't flag it as "accurately an integer" if the number
2090 came from a (by definition imprecise) NV operation, and
2091 we're outside the range of NV integer precision */
2092#endif
a43d94f2 2093 && SvNOK(sv)
28e5dec8
JH
2094 )
2095 SvIOK_on(sv);
25da4f38 2096 SvIsUV_on(sv);
1c846c1f 2097 DEBUG_c(PerlIO_printf(Perl_debug_log,
57def98f 2098 "0x%"UVxf" 2iv(%"UVuf" => %"IVdf") (as unsigned)\n",
56431972 2099 PTR2UV(sv),
57def98f
JH
2100 SvUVX(sv),
2101 SvUVX(sv)));
25da4f38 2102 }
748a9306
LW
2103 }
2104 else if (SvPOKp(sv) && SvLEN(sv)) {
c2988b20 2105 UV value;
504618e9 2106 const int numtype = grok_number(SvPVX_const(sv), SvCUR(sv), &value);
af359546 2107 /* We want to avoid a possible problem when we cache an IV/ a UV which
25da4f38 2108 may be later translated to an NV, and the resulting NV is not
c2988b20
NC
2109 the same as the direct translation of the initial string
2110 (eg 123.456 can shortcut to the IV 123 with atol(), but we must
2111 be careful to ensure that the value with the .456 is around if the
2112 NV value is requested in the future).
1c846c1f 2113
af359546 2114 This means that if we cache such an IV/a UV, we need to cache the
25da4f38 2115 NV as well. Moreover, we trade speed for space, and do not
28e5dec8 2116 cache the NV if we are sure it's not needed.
25da4f38 2117 */
16b7a9a4 2118
c2988b20
NC
2119 /* SVt_PVNV is one higher than SVt_PVIV, hence this order */
2120 if ((numtype & (IS_NUMBER_IN_UV | IS_NUMBER_NOT_INT))
2121 == IS_NUMBER_IN_UV) {
5e045b90 2122 /* It's definitely an integer, only upgrade to PVIV */
28e5dec8
JH
2123 if (SvTYPE(sv) < SVt_PVIV)
2124 sv_upgrade(sv, SVt_PVIV);
f7bbb42a 2125 (void)SvIOK_on(sv);
c2988b20
NC
2126 } else if (SvTYPE(sv) < SVt_PVNV)
2127 sv_upgrade(sv, SVt_PVNV);
28e5dec8 2128
f2524eef 2129 /* If NVs preserve UVs then we only use the UV value if we know that
c2988b20
NC
2130 we aren't going to call atof() below. If NVs don't preserve UVs
2131 then the value returned may have more precision than atof() will
2132 return, even though value isn't perfectly accurate. */
2133 if ((numtype & (IS_NUMBER_IN_UV
2134#ifdef NV_PRESERVES_UV
2135 | IS_NUMBER_NOT_INT
2136#endif
2137 )) == IS_NUMBER_IN_UV) {
2138 /* This won't turn off the public IOK flag if it was set above */
2139 (void)SvIOKp_on(sv);
2140
2141 if (!(numtype & IS_NUMBER_NEG)) {
2142 /* positive */;
2143 if (value <= (UV)IV_MAX) {
45977657 2144 SvIV_set(sv, (IV)value);
c2988b20 2145 } else {
af359546 2146 /* it didn't overflow, and it was positive. */
607fa7f2 2147 SvUV_set(sv, value);
c2988b20
NC
2148 SvIsUV_on(sv);
2149 }
2150 } else {
2151 /* 2s complement assumption */
2152 if (value <= (UV)IV_MIN) {
45977657 2153 SvIV_set(sv, -(IV)value);
c2988b20
NC
2154 } else {
2155 /* Too negative for an IV. This is a double upgrade, but
d1be9408 2156 I'm assuming it will be rare. */
c2988b20
NC
2157 if (SvTYPE(sv) < SVt_PVNV)
2158 sv_upgrade(sv, SVt_PVNV);
2159 SvNOK_on(sv);
2160 SvIOK_off(sv);
2161 SvIOKp_on(sv);
9d6ce603 2162 SvNV_set(sv, -(NV)value);
45977657 2163 SvIV_set(sv, IV_MIN);
c2988b20
NC
2164 }
2165 }
2166 }
2167 /* For !NV_PRESERVES_UV and IS_NUMBER_IN_UV and IS_NUMBER_NOT_INT we
2168 will be in the previous block to set the IV slot, and the next
2169 block to set the NV slot. So no else here. */
2170
2171 if ((numtype & (IS_NUMBER_IN_UV | IS_NUMBER_NOT_INT))
2172 != IS_NUMBER_IN_UV) {
2173 /* It wasn't an (integer that doesn't overflow the UV). */
3f7c398e 2174 SvNV_set(sv, Atof(SvPVX_const(sv)));
28e5dec8 2175
c2988b20
NC
2176 if (! numtype && ckWARN(WARN_NUMERIC))
2177 not_a_number(sv);
28e5dec8 2178
65202027 2179#if defined(USE_LONG_DOUBLE)
c2988b20
NC
2180 DEBUG_c(PerlIO_printf(Perl_debug_log, "0x%"UVxf" 2iv(%" PERL_PRIgldbl ")\n",
2181 PTR2UV(sv), SvNVX(sv)));
65202027 2182#else
1779d84d 2183 DEBUG_c(PerlIO_printf(Perl_debug_log, "0x%"UVxf" 2iv(%"NVgf")\n",
c2988b20 2184 PTR2UV(sv), SvNVX(sv)));
65202027 2185#endif
28e5dec8 2186
28e5dec8 2187#ifdef NV_PRESERVES_UV
af359546
NC
2188 (void)SvIOKp_on(sv);
2189 (void)SvNOK_on(sv);
2190 if (SvNVX(sv) < (NV)IV_MAX + 0.5) {
2191 SvIV_set(sv, I_V(SvNVX(sv)));
2192 if ((NV)(SvIVX(sv)) == SvNVX(sv)) {
2193 SvIOK_on(sv);
2194 } else {
6f207bd3 2195 NOOP; /* Integer is imprecise. NOK, IOKp */
af359546
NC
2196 }
2197 /* UV will not work better than IV */
2198 } else {
2199 if (SvNVX(sv) > (NV)UV_MAX) {
2200 SvIsUV_on(sv);
2201 /* Integer is inaccurate. NOK, IOKp, is UV */
2202 SvUV_set(sv, UV_MAX);
af359546
NC
2203 } else {
2204 SvUV_set(sv, U_V(SvNVX(sv)));
2205 /* 0xFFFFFFFFFFFFFFFF not an issue in here, NVs
2206 NV preservse UV so can do correct comparison. */
2207 if ((NV)(SvUVX(sv)) == SvNVX(sv)) {
2208 SvIOK_on(sv);
af359546 2209 } else {
6f207bd3 2210 NOOP; /* Integer is imprecise. NOK, IOKp, is UV */
af359546
NC
2211 }
2212 }
4b0c9573 2213 SvIsUV_on(sv);
af359546 2214 }
28e5dec8 2215#else /* NV_PRESERVES_UV */
c2988b20
NC
2216 if ((numtype & (IS_NUMBER_IN_UV | IS_NUMBER_NOT_INT))
2217 == (IS_NUMBER_IN_UV | IS_NUMBER_NOT_INT)) {
af359546 2218 /* The IV/UV slot will have been set from value returned by
c2988b20
NC
2219 grok_number above. The NV slot has just been set using
2220 Atof. */
560b0c46 2221 SvNOK_on(sv);
c2988b20
NC
2222 assert (SvIOKp(sv));
2223 } else {
2224 if (((UV)1 << NV_PRESERVES_UV_BITS) >
2225 U_V(SvNVX(sv) > 0 ? SvNVX(sv) : -SvNVX(sv))) {
2226 /* Small enough to preserve all bits. */
2227 (void)SvIOKp_on(sv);
2228 SvNOK_on(sv);
45977657 2229 SvIV_set(sv, I_V(SvNVX(sv)));
c2988b20
NC
2230 if ((NV)(SvIVX(sv)) == SvNVX(sv))
2231 SvIOK_on(sv);
2232 /* Assumption: first non-preserved integer is < IV_MAX,
2233 this NV is in the preserved range, therefore: */
2234 if (!(U_V(SvNVX(sv) > 0 ? SvNVX(sv) : -SvNVX(sv))
2235 < (UV)IV_MAX)) {
32fdb065 2236 Perl_croak(aTHX_ "sv_2iv assumed (U_V(fabs((double)SvNVX(sv))) < (UV)IV_MAX) but SvNVX(sv)=%"NVgf" U_V is 0x%"UVxf", IV_MAX is 0x%"UVxf"\n", SvNVX(sv), U_V(SvNVX(sv)), (UV)IV_MAX);
c2988b20
NC
2237 }
2238 } else {
2239 /* IN_UV NOT_INT
2240 0 0 already failed to read UV.
2241 0 1 already failed to read UV.
2242 1 0 you won't get here in this case. IV/UV
2243 slot set, public IOK, Atof() unneeded.
2244 1 1 already read UV.
2245 so there's no point in sv_2iuv_non_preserve() attempting
2246 to use atol, strtol, strtoul etc. */
47031da6 2247# ifdef DEBUGGING
40a17c4c 2248 sv_2iuv_non_preserve (sv, numtype);
47031da6
NC
2249# else
2250 sv_2iuv_non_preserve (sv);
2251# endif
c2988b20
NC
2252 }
2253 }
28e5dec8 2254#endif /* NV_PRESERVES_UV */
a43d94f2
NC
2255 /* It might be more code efficient to go through the entire logic above
2256 and conditionally set with SvIOKp_on() rather than SvIOK(), but it
2257 gets complex and potentially buggy, so more programmer efficient
2258 to do it this way, by turning off the public flags: */
2259 if (!numtype)
2260 SvFLAGS(sv) &= ~(SVf_IOK|SVf_NOK);
25da4f38 2261 }
af359546
NC
2262 }
2263 else {
675c862f 2264 if (isGV_with_GP(sv))
159b6efe 2265 return glob_2number(MUTABLE_GV(sv));
180488f8 2266
af359546
NC
2267 if (!(SvFLAGS(sv) & SVs_PADTMP)) {
2268 if (!PL_localizing && ckWARN(WARN_UNINITIALIZED))
2269 report_uninit(sv);
2270 }
25da4f38
IZ
2271 if (SvTYPE(sv) < SVt_IV)
2272 /* Typically the caller expects that sv_any is not NULL now. */
2273 sv_upgrade(sv, SVt_IV);
af359546
NC
2274 /* Return 0 from the caller. */
2275 return TRUE;
2276 }
2277 return FALSE;
2278}
2279
2280/*
2281=for apidoc sv_2iv_flags
2282
2283Return the integer value of an SV, doing any necessary string
2284conversion. If flags includes SV_GMAGIC, does an mg_get() first.
2285Normally used via the C<SvIV(sv)> and C<SvIVx(sv)> macros.
2286
2287=cut
2288*/
2289
2290IV
5de3775c 2291Perl_sv_2iv_flags(pTHX_ register SV *const sv, const I32 flags)
af359546 2292{
97aff369 2293 dVAR;
af359546 2294 if (!sv)
a0d0e21e 2295 return 0;
cecf5685
NC
2296 if (SvGMAGICAL(sv) || (SvTYPE(sv) == SVt_PVGV && SvVALID(sv))) {
2297 /* FBMs use the same flag bit as SVf_IVisUV, so must let them
50caf62e
NC
2298 cache IVs just in case. In practice it seems that they never
2299 actually anywhere accessible by user Perl code, let alone get used
2300 in anything other than a string context. */
af359546
NC
2301 if (flags & SV_GMAGIC)
2302 mg_get(sv);
2303 if (SvIOKp(sv))
2304 return SvIVX(sv);
2305 if (SvNOKp(sv)) {
2306 return I_V(SvNVX(sv));
2307 }
71c558c3
NC
2308 if (SvPOKp(sv) && SvLEN(sv)) {
2309 UV value;
2310 const int numtype
2311 = grok_number(SvPVX_const(sv), SvCUR(sv), &value);
2312
2313 if ((numtype & (IS_NUMBER_IN_UV | IS_NUMBER_NOT_INT))
2314 == IS_NUMBER_IN_UV) {
2315 /* It's definitely an integer */
2316 if (numtype & IS_NUMBER_NEG) {
2317 if (value < (UV)IV_MIN)
2318 return -(IV)value;
2319 } else {
2320 if (value < (UV)IV_MAX)
2321 return (IV)value;
2322 }
2323 }
2324 if (!numtype) {
2325 if (ckWARN(WARN_NUMERIC))
2326 not_a_number(sv);
2327 }
2328 return I_V(Atof(SvPVX_const(sv)));
2329 }
1c7ff15e
NC
2330 if (SvROK(sv)) {
2331 goto return_rok;
af359546 2332 }
1c7ff15e
NC
2333 assert(SvTYPE(sv) >= SVt_PVMG);
2334 /* This falls through to the report_uninit inside S_sv_2iuv_common. */
4cb1ec55 2335 } else if (SvTHINKFIRST(sv)) {
af359546 2336 if (SvROK(sv)) {
1c7ff15e 2337 return_rok:
af359546
NC
2338 if (SvAMAGIC(sv)) {
2339 SV * const tmpstr=AMG_CALLun(sv,numer);
2340 if (tmpstr && (!SvROK(tmpstr) || (SvRV(tmpstr) != SvRV(sv)))) {
2341 return SvIV(tmpstr);
2342 }
2343 }
2344 return PTR2IV(SvRV(sv));
2345 }
2346 if (SvIsCOW(sv)) {
2347 sv_force_normal_flags(sv, 0);
2348 }
2349 if (SvREADONLY(sv) && !SvOK(sv)) {
2350 if (ckWARN(WARN_UNINITIALIZED))
2351 report_uninit(sv);
2352 return 0;
2353 }
2354 }
2355 if (!SvIOKp(sv)) {
2356 if (S_sv_2iuv_common(aTHX_ sv))
2357 return 0;
79072805 2358 }
1d7c1841
GS
2359 DEBUG_c(PerlIO_printf(Perl_debug_log, "0x%"UVxf" 2iv(%"IVdf")\n",
2360 PTR2UV(sv),SvIVX(sv)));
25da4f38 2361 return SvIsUV(sv) ? (IV)SvUVX(sv) : SvIVX(sv);
79072805
LW
2362}
2363
645c22ef 2364/*
891f9566 2365=for apidoc sv_2uv_flags
645c22ef
DM
2366
2367Return the unsigned integer value of an SV, doing any necessary string
891f9566
YST
2368conversion. If flags includes SV_GMAGIC, does an mg_get() first.
2369Normally used via the C<SvUV(sv)> and C<SvUVx(sv)> macros.
645c22ef
DM
2370
2371=cut
2372*/
2373
ff68c719 2374UV
5de3775c 2375Perl_sv_2uv_flags(pTHX_ register SV *const sv, const I32 flags)
ff68c719 2376{
97aff369 2377 dVAR;
ff68c719 2378 if (!sv)
2379 return 0;
cecf5685
NC
2380 if (SvGMAGICAL(sv) || (SvTYPE(sv) == SVt_PVGV && SvVALID(sv))) {
2381 /* FBMs use the same flag bit as SVf_IVisUV, so must let them
50caf62e 2382 cache IVs just in case. */
891f9566
YST
2383 if (flags & SV_GMAGIC)
2384 mg_get(sv);
ff68c719 2385 if (SvIOKp(sv))
2386 return SvUVX(sv);
2387 if (SvNOKp(sv))
2388 return U_V(SvNVX(sv));
71c558c3
NC
2389 if (SvPOKp(sv) && SvLEN(sv)) {
2390 UV value;
2391 const int numtype
2392 = grok_number(SvPVX_const(sv), SvCUR(sv), &value);
2393
2394 if ((numtype & (IS_NUMBER_IN_UV | IS_NUMBER_NOT_INT))
2395 == IS_NUMBER_IN_UV) {
2396 /* It's definitely an integer */
2397 if (!(numtype & IS_NUMBER_NEG))
2398 return value;
2399 }
2400 if (!numtype) {
2401 if (ckWARN(WARN_NUMERIC))
2402 not_a_number(sv);
2403 }
2404 return U_V(Atof(SvPVX_const(sv)));
2405 }
1c7ff15e
NC
2406 if (SvROK(sv)) {
2407 goto return_rok;
3fe9a6f1 2408 }
1c7ff15e
NC
2409 assert(SvTYPE(sv) >= SVt_PVMG);
2410 /* This falls through to the report_uninit inside S_sv_2iuv_common. */
4cb1ec55 2411 } else if (SvTHINKFIRST(sv)) {
ff68c719 2412 if (SvROK(sv)) {
1c7ff15e 2413 return_rok:
deb46114
NC
2414 if (SvAMAGIC(sv)) {
2415 SV *const tmpstr = AMG_CALLun(sv,numer);
2416 if (tmpstr && (!SvROK(tmpstr) || (SvRV(tmpstr) != SvRV(sv)))) {
2417 return SvUV(tmpstr);
2418 }
2419 }
2420 return PTR2UV(SvRV(sv));
ff68c719 2421 }
765f542d
NC
2422 if (SvIsCOW(sv)) {
2423 sv_force_normal_flags(sv, 0);
8a818333 2424 }
0336b60e 2425 if (SvREADONLY(sv) && !SvOK(sv)) {
0336b60e 2426 if (ckWARN(WARN_UNINITIALIZED))
29489e7c 2427 report_uninit(sv);
ff68c719 2428 return 0;
2429 }
2430 }
af359546
NC
2431 if (!SvIOKp(sv)) {
2432 if (S_sv_2iuv_common(aTHX_ sv))
2433 return 0;
ff68c719 2434 }
25da4f38 2435
1d7c1841
GS
2436 DEBUG_c(PerlIO_printf(Perl_debug_log, "0x%"UVxf" 2uv(%"UVuf")\n",
2437 PTR2UV(sv),SvUVX(sv)));
25da4f38 2438 return SvIsUV(sv) ? SvUVX(sv) : (UV)SvIVX(sv);
ff68c719 2439}
2440
645c22ef
DM
2441/*
2442=for apidoc sv_2nv
2443
2444Return the num value of an SV, doing any necessary string or integer
2445conversion, magic etc. Normally used via the C<SvNV(sv)> and C<SvNVx(sv)>
2446macros.
2447
2448=cut
2449*/
2450
65202027 2451NV
5de3775c 2452Perl_sv_2nv(pTHX_ register SV *const sv)
79072805 2453{
97aff369 2454 dVAR;
79072805
LW
2455 if (!sv)
2456 return 0.0;
cecf5685
NC
2457 if (SvGMAGICAL(sv) || (SvTYPE(sv) == SVt_PVGV && SvVALID(sv))) {
2458 /* FBMs use the same flag bit as SVf_IVisUV, so must let them
50caf62e 2459 cache IVs just in case. */
463ee0b2
LW
2460 mg_get(sv);
2461 if (SvNOKp(sv))
2462 return SvNVX(sv);
0aa395f8 2463 if ((SvPOKp(sv) && SvLEN(sv)) && !SvIOKp(sv)) {
041457d9 2464 if (!SvIOKp(sv) && ckWARN(WARN_NUMERIC) &&
504618e9 2465 !grok_number(SvPVX_const(sv), SvCUR(sv), NULL))
a0d0e21e 2466 not_a_number(sv);
3f7c398e 2467 return Atof(SvPVX_const(sv));
a0d0e21e 2468 }
25da4f38 2469 if (SvIOKp(sv)) {
1c846c1f 2470 if (SvIsUV(sv))
65202027 2471 return (NV)SvUVX(sv);
25da4f38 2472 else
65202027 2473 return (NV)SvIVX(sv);
47a72cb8
NC
2474 }
2475 if (SvROK(sv)) {
2476 goto return_rok;
2477 }
2478 assert(SvTYPE(sv) >= SVt_PVMG);
2479 /* This falls through to the report_uninit near the end of the
2480 function. */
2481 } else if (SvTHINKFIRST(sv)) {
a0d0e21e 2482 if (SvROK(sv)) {
47a72cb8 2483 return_rok:
deb46114
NC
2484 if (SvAMAGIC(sv)) {
2485 SV *const tmpstr = AMG_CALLun(sv,numer);
2486 if (tmpstr && (!SvROK(tmpstr) || (SvRV(tmpstr) != SvRV(sv)))) {
2487 return SvNV(tmpstr);
2488 }
2489 }
2490 return PTR2NV(SvRV(sv));
a0d0e21e 2491 }
765f542d
NC
2492 if (SvIsCOW(sv)) {
2493 sv_force_normal_flags(sv, 0);
8a818333 2494 }
0336b60e 2495 if (SvREADONLY(sv) && !SvOK(sv)) {
599cee73 2496 if (ckWARN(WARN_UNINITIALIZED))
29489e7c 2497 report_uninit(sv);
ed6116ce
LW
2498 return 0.0;
2499 }
79072805
LW
2500 }
2501 if (SvTYPE(sv) < SVt_NV) {
7e25a7e9
NC
2502 /* The logic to use SVt_PVNV if necessary is in sv_upgrade. */
2503 sv_upgrade(sv, SVt_NV);
906f284f 2504#ifdef USE_LONG_DOUBLE
097ee67d 2505 DEBUG_c({
f93f4e46 2506 STORE_NUMERIC_LOCAL_SET_STANDARD();
1d7c1841
GS
2507 PerlIO_printf(Perl_debug_log,
2508 "0x%"UVxf" num(%" PERL_PRIgldbl ")\n",
2509 PTR2UV(sv), SvNVX(sv));
572bbb43
GS
2510 RESTORE_NUMERIC_LOCAL();
2511 });
65202027 2512#else
572bbb43 2513 DEBUG_c({
f93f4e46 2514 STORE_NUMERIC_LOCAL_SET_STANDARD();
1779d84d 2515 PerlIO_printf(Perl_debug_log, "0x%"UVxf" num(%"NVgf")\n",
1d7c1841 2516 PTR2UV(sv), SvNVX(sv));
097ee67d
JH
2517 RESTORE_NUMERIC_LOCAL();
2518 });
572bbb43 2519#endif
79072805
LW
2520 }
2521 else if (SvTYPE(sv) < SVt_PVNV)
2522 sv_upgrade(sv, SVt_PVNV);
59d8ce62
NC
2523 if (SvNOKp(sv)) {
2524 return SvNVX(sv);
61604483 2525 }
59d8ce62 2526 if (SvIOKp(sv)) {
9d6ce603 2527 SvNV_set(sv, SvIsUV(sv) ? (NV)SvUVX(sv) : (NV)SvIVX(sv));
28e5dec8 2528#ifdef NV_PRESERVES_UV
a43d94f2
NC
2529 if (SvIOK(sv))
2530 SvNOK_on(sv);
2531 else
2532 SvNOKp_on(sv);
28e5dec8
JH
2533#else
2534 /* Only set the public NV OK flag if this NV preserves the IV */
2535 /* Check it's not 0xFFFFFFFFFFFFFFFF */
a43d94f2
NC
2536 if (SvIOK(sv) &&
2537 SvIsUV(sv) ? ((SvUVX(sv) != UV_MAX)&&(SvUVX(sv) == U_V(SvNVX(sv))))
28e5dec8
JH
2538 : (SvIVX(sv) == I_V(SvNVX(sv))))
2539 SvNOK_on(sv);
2540 else
2541 SvNOKp_on(sv);
2542#endif
93a17b20 2543 }
748a9306 2544 else if (SvPOKp(sv) && SvLEN(sv)) {
c2988b20 2545 UV value;
3f7c398e 2546 const int numtype = grok_number(SvPVX_const(sv), SvCUR(sv), &value);
041457d9 2547 if (!SvIOKp(sv) && !numtype && ckWARN(WARN_NUMERIC))
a0d0e21e 2548 not_a_number(sv);
28e5dec8 2549#ifdef NV_PRESERVES_UV
c2988b20
NC
2550 if ((numtype & (IS_NUMBER_IN_UV | IS_NUMBER_NOT_INT))
2551 == IS_NUMBER_IN_UV) {
5e045b90 2552 /* It's definitely an integer */
9d6ce603 2553 SvNV_set(sv, (numtype & IS_NUMBER_NEG) ? -(NV)value : (NV)value);
c2988b20 2554 } else
3f7c398e 2555 SvNV_set(sv, Atof(SvPVX_const(sv)));
a43d94f2
NC
2556 if (numtype)
2557 SvNOK_on(sv);
2558 else
2559 SvNOKp_on(sv);
28e5dec8 2560#else
3f7c398e 2561 SvNV_set(sv, Atof(SvPVX_const(sv)));
28e5dec8
JH
2562 /* Only set the public NV OK flag if this NV preserves the value in
2563 the PV at least as well as an IV/UV would.
2564 Not sure how to do this 100% reliably. */
2565 /* if that shift count is out of range then Configure's test is
2566 wonky. We shouldn't be in here with NV_PRESERVES_UV_BITS ==
2567 UV_BITS */
2568 if (((UV)1 << NV_PRESERVES_UV_BITS) >
c2988b20 2569 U_V(SvNVX(sv) > 0 ? SvNVX(sv) : -SvNVX(sv))) {
28e5dec8 2570 SvNOK_on(sv); /* Definitely small enough to preserve all bits */
c2988b20
NC
2571 } else if (!(numtype & IS_NUMBER_IN_UV)) {
2572 /* Can't use strtol etc to convert this string, so don't try.
2573 sv_2iv and sv_2uv will use the NV to convert, not the PV. */
2574 SvNOK_on(sv);
2575 } else {
2576 /* value has been set. It may not be precise. */
2577 if ((numtype & IS_NUMBER_NEG) && (value > (UV)IV_MIN)) {
2578 /* 2s complement assumption for (UV)IV_MIN */
2579 SvNOK_on(sv); /* Integer is too negative. */
2580 } else {
2581 SvNOKp_on(sv);
2582 SvIOKp_on(sv);
6fa402ec 2583
c2988b20 2584 if (numtype & IS_NUMBER_NEG) {
45977657 2585 SvIV_set(sv, -(IV)value);
c2988b20 2586 } else if (value <= (UV)IV_MAX) {
45977657 2587 SvIV_set(sv, (IV)value);
c2988b20 2588 } else {
607fa7f2 2589 SvUV_set(sv, value);
c2988b20
NC
2590 SvIsUV_on(sv);
2591 }
2592
2593 if (numtype & IS_NUMBER_NOT_INT) {
2594 /* I believe that even if the original PV had decimals,
2595 they are lost beyond the limit of the FP precision.
2596 However, neither is canonical, so both only get p
2597 flags. NWC, 2000/11/25 */
2598 /* Both already have p flags, so do nothing */
2599 } else {
66a1b24b 2600 const NV nv = SvNVX(sv);
c2988b20
NC
2601 if (SvNVX(sv) < (NV)IV_MAX + 0.5) {
2602 if (SvIVX(sv) == I_V(nv)) {
2603 SvNOK_on(sv);
c2988b20 2604 } else {
c2988b20
NC
2605 /* It had no "." so it must be integer. */
2606 }
00b6aa41 2607 SvIOK_on(sv);
c2988b20
NC
2608 } else {
2609 /* between IV_MAX and NV(UV_MAX).
2610 Could be slightly > UV_MAX */
6fa402ec 2611
c2988b20
NC
2612 if (numtype & IS_NUMBER_NOT_INT) {
2613 /* UV and NV both imprecise. */
2614 } else {
66a1b24b 2615 const UV nv_as_uv = U_V(nv);
c2988b20
NC
2616
2617 if (value == nv_as_uv && SvUVX(sv) != UV_MAX) {
2618 SvNOK_on(sv);
c2988b20 2619 }
00b6aa41 2620 SvIOK_on(sv);
c2988b20
NC
2621 }
2622 }
2623 }
2624 }
2625 }
a43d94f2
NC
2626 /* It might be more code efficient to go through the entire logic above
2627 and conditionally set with SvNOKp_on() rather than SvNOK(), but it
2628 gets complex and potentially buggy, so more programmer efficient
2629 to do it this way, by turning off the public flags: */
2630 if (!numtype)
2631 SvFLAGS(sv) &= ~(SVf_IOK|SVf_NOK);
28e5dec8 2632#endif /* NV_PRESERVES_UV */
93a17b20 2633 }
79072805 2634 else {
f7877b28 2635 if (isGV_with_GP(sv)) {
159b6efe 2636 glob_2number(MUTABLE_GV(sv));
180488f8
NC
2637 return 0.0;
2638 }
2639
041457d9 2640 if (!PL_localizing && !(SvFLAGS(sv) & SVs_PADTMP) && ckWARN(WARN_UNINITIALIZED))
29489e7c 2641 report_uninit(sv);
7e25a7e9
NC
2642 assert (SvTYPE(sv) >= SVt_NV);
2643 /* Typically the caller expects that sv_any is not NULL now. */
2644 /* XXX Ilya implies that this is a bug in callers that assume this
2645 and ideally should be fixed. */
a0d0e21e 2646 return 0.0;
79072805 2647 }
572bbb43 2648#if defined(USE_LONG_DOUBLE)
097ee67d 2649 DEBUG_c({
f93f4e46 2650 STORE_NUMERIC_LOCAL_SET_STANDARD();
1d7c1841
GS
2651 PerlIO_printf(Perl_debug_log, "0x%"UVxf" 2nv(%" PERL_PRIgldbl ")\n",
2652 PTR2UV(sv), SvNVX(sv));
572bbb43
GS
2653 RESTORE_NUMERIC_LOCAL();
2654 });
65202027 2655#else
572bbb43 2656 DEBUG_c({
f93f4e46 2657 STORE_NUMERIC_LOCAL_SET_STANDARD();
1779d84d 2658 PerlIO_printf(Perl_debug_log, "0x%"UVxf" 1nv(%"NVgf")\n",
1d7c1841 2659 PTR2UV(sv), SvNVX(sv));
097ee67d
JH
2660 RESTORE_NUMERIC_LOCAL();
2661 });
572bbb43 2662#endif
463ee0b2 2663 return SvNVX(sv);
79072805
LW
2664}
2665
800401ee
JH
2666/*
2667=for apidoc sv_2num
2668
2669Return an SV with the numeric value of the source SV, doing any necessary
a196a5fa
JH
2670reference or overload conversion. You must use the C<SvNUM(sv)> macro to
2671access this function.
800401ee
JH
2672
2673=cut
2674*/
2675
2676SV *
5de3775c 2677Perl_sv_2num(pTHX_ register SV *const sv)
800401ee 2678{
7918f24d
NC
2679 PERL_ARGS_ASSERT_SV_2NUM;
2680
b9ee0594
RGS
2681 if (!SvROK(sv))
2682 return sv;
800401ee
JH
2683 if (SvAMAGIC(sv)) {
2684 SV * const tmpsv = AMG_CALLun(sv,numer);
2685 if (tmpsv && (!SvROK(tmpsv) || (SvRV(tmpsv) != SvRV(sv))))
2686 return sv_2num(tmpsv);
2687 }
2688 return sv_2mortal(newSVuv(PTR2UV(SvRV(sv))));
2689}
2690
645c22ef
DM
2691/* uiv_2buf(): private routine for use by sv_2pv_flags(): print an IV or
2692 * UV as a string towards the end of buf, and return pointers to start and
2693 * end of it.
2694 *
2695 * We assume that buf is at least TYPE_CHARS(UV) long.
2696 */
2697
864dbfa3 2698static char *
5de3775c 2699S_uiv_2buf(char *const buf, const IV iv, UV uv, const int is_uv, char **const peob)
25da4f38 2700{
25da4f38 2701 char *ptr = buf + TYPE_CHARS(UV);
823a54a3 2702 char * const ebuf = ptr;
25da4f38 2703 int sign;
25da4f38 2704
7918f24d
NC
2705 PERL_ARGS_ASSERT_UIV_2BUF;
2706
25da4f38
IZ
2707 if (is_uv)
2708 sign = 0;
2709 else if (iv >= 0) {
2710 uv = iv;
2711 sign = 0;
2712 } else {
2713 uv = -iv;
2714 sign = 1;
2715 }
2716 do {
eb160463 2717 *--ptr = '0' + (char)(uv % 10);
25da4f38
IZ
2718 } while (uv /= 10);
2719 if (sign)
2720 *--ptr = '-';
2721 *peob = ebuf;
2722 return ptr;
2723}
2724
645c22ef
DM
2725/*
2726=for apidoc sv_2pv_flags
2727
ff276b08 2728Returns a pointer to the string value of an SV, and sets *lp to its length.
645c22ef
DM
2729If flags includes SV_GMAGIC, does an mg_get() first. Coerces sv to a string
2730if necessary.
2731Normally invoked via the C<SvPV_flags> macro. C<sv_2pv()> and C<sv_2pv_nomg>
2732usually end up here too.
2733
2734=cut
2735*/
2736
8d6d96c1 2737char *
5de3775c 2738Perl_sv_2pv_flags(pTHX_ register SV *const sv, STRLEN *const lp, const I32 flags)
8d6d96c1 2739{
97aff369 2740 dVAR;
79072805 2741 register char *s;
79072805 2742
463ee0b2 2743 if (!sv) {
cdb061a3
NC
2744 if (lp)
2745 *lp = 0;
73d840c0 2746 return (char *)"";
463ee0b2 2747 }
8990e307 2748 if (SvGMAGICAL(sv)) {
8d6d96c1
HS
2749 if (flags & SV_GMAGIC)
2750 mg_get(sv);
463ee0b2 2751 if (SvPOKp(sv)) {
cdb061a3
NC
2752 if (lp)
2753 *lp = SvCUR(sv);
10516c54
NC
2754 if (flags & SV_MUTABLE_RETURN)
2755 return SvPVX_mutable(sv);
4d84ee25
NC
2756 if (flags & SV_CONST_RETURN)
2757 return (char *)SvPVX_const(sv);
463ee0b2
LW
2758 return SvPVX(sv);
2759 }
75dfc8ec
NC
2760 if (SvIOKp(sv) || SvNOKp(sv)) {
2761 char tbuf[64]; /* Must fit sprintf/Gconvert of longest IV/NV */
75dfc8ec
NC
2762 STRLEN len;
2763
2764 if (SvIOKp(sv)) {
e80fed9d 2765 len = SvIsUV(sv)
d9fad198
JH
2766 ? my_snprintf(tbuf, sizeof(tbuf), "%"UVuf, (UV)SvUVX(sv))
2767 : my_snprintf(tbuf, sizeof(tbuf), "%"IVdf, (IV)SvIVX(sv));
75dfc8ec 2768 } else {
e8ada2d0
NC
2769 Gconvert(SvNVX(sv), NV_DIG, 0, tbuf);
2770 len = strlen(tbuf);
75dfc8ec 2771 }
b5b886f0
NC
2772 assert(!SvROK(sv));
2773 {
75dfc8ec
NC
2774 dVAR;
2775
2776#ifdef FIXNEGATIVEZERO
e8ada2d0
NC
2777 if (len == 2 && tbuf[0] == '-' && tbuf[1] == '0') {
2778 tbuf[0] = '0';
2779 tbuf[1] = 0;
75dfc8ec
NC
2780 len = 1;
2781 }
2782#endif
2783 SvUPGRADE(sv, SVt_PV);
2784 if (lp)
2785 *lp = len;
2786 s = SvGROW_mutable(sv, len + 1);
2787 SvCUR_set(sv, len);
2788 SvPOKp_on(sv);
10edeb5d 2789 return (char*)memcpy(s, tbuf, len + 1);
75dfc8ec 2790 }
463ee0b2 2791 }
1c7ff15e
NC
2792 if (SvROK(sv)) {
2793 goto return_rok;
2794 }
2795 assert(SvTYPE(sv) >= SVt_PVMG);
2796 /* This falls through to the report_uninit near the end of the
2797 function. */
2798 } else if (SvTHINKFIRST(sv)) {
ed6116ce 2799 if (SvROK(sv)) {
1c7ff15e 2800 return_rok:
deb46114
NC
2801 if (SvAMAGIC(sv)) {
2802 SV *const tmpstr = AMG_CALLun(sv,string);
2803 if (tmpstr && (!SvROK(tmpstr) || (SvRV(tmpstr) != SvRV(sv)))) {
2804 /* Unwrap this: */
2805 /* char *pv = lp ? SvPV(tmpstr, *lp) : SvPV_nolen(tmpstr);
2806 */
2807
2808 char *pv;
2809 if ((SvFLAGS(tmpstr) & (SVf_POK)) == SVf_POK) {
2810 if (flags & SV_CONST_RETURN) {
2811 pv = (char *) SvPVX_const(tmpstr);
2812 } else {
2813 pv = (flags & SV_MUTABLE_RETURN)
2814 ? SvPVX_mutable(tmpstr) : SvPVX(tmpstr);
2815 }
2816 if (lp)
2817 *lp = SvCUR(tmpstr);
50adf7d2 2818 } else {
deb46114 2819 pv = sv_2pv_flags(tmpstr, lp, flags);
50adf7d2 2820 }
deb46114
NC
2821 if (SvUTF8(tmpstr))
2822 SvUTF8_on(sv);
2823 else
2824 SvUTF8_off(sv);
2825 return pv;
50adf7d2 2826 }
deb46114
NC
2827 }
2828 {
fafee734
NC
2829 STRLEN len;
2830 char *retval;
2831 char *buffer;
d2c6dc5e 2832 SV *const referent = SvRV(sv);
d8eae41e
NC
2833
2834 if (!referent) {
fafee734
NC
2835 len = 7;
2836 retval = buffer = savepvn("NULLREF", len);
5c35adbb 2837 } else if (SvTYPE(referent) == SVt_REGEXP) {
d2c6dc5e 2838 REGEXP * const re = (REGEXP *)MUTABLE_PTR(referent);
67d2d14d
AB
2839 I32 seen_evals = 0;
2840
2841 assert(re);
2842
2843 /* If the regex is UTF-8 we want the containing scalar to
2844 have an UTF-8 flag too */
2845 if (RX_UTF8(re))
2846 SvUTF8_on(sv);
2847 else
2848 SvUTF8_off(sv);
2849
2850 if ((seen_evals = RX_SEEN_EVALS(re)))
2851 PL_reginterp_cnt += seen_evals;
2852
2853 if (lp)
2854 *lp = RX_WRAPLEN(re);
2855
2856 return RX_WRAPPED(re);
d8eae41e
NC
2857 } else {
2858 const char *const typestr = sv_reftype(referent, 0);
fafee734
NC
2859 const STRLEN typelen = strlen(typestr);
2860 UV addr = PTR2UV(referent);
2861 const char *stashname = NULL;
2862 STRLEN stashnamelen = 0; /* hush, gcc */
2863 const char *buffer_end;
d8eae41e 2864
d8eae41e 2865 if (SvOBJECT(referent)) {
fafee734
NC
2866 const HEK *const name = HvNAME_HEK(SvSTASH(referent));
2867
2868 if (name) {
2869 stashname = HEK_KEY(name);
2870 stashnamelen = HEK_LEN(name);
2871
2872 if (HEK_UTF8(name)) {
2873 SvUTF8_on(sv);
2874 } else {
2875 SvUTF8_off(sv);
2876 }
2877 } else {
2878 stashname = "__ANON__";
2879 stashnamelen = 8;
2880 }
2881 len = stashnamelen + 1 /* = */ + typelen + 3 /* (0x */
2882 + 2 * sizeof(UV) + 2 /* )\0 */;
2883 } else {
2884 len = typelen + 3 /* (0x */
2885 + 2 * sizeof(UV) + 2 /* )\0 */;
d8eae41e 2886 }
fafee734
NC
2887
2888 Newx(buffer, len, char);
2889 buffer_end = retval = buffer + len;
2890
2891 /* Working backwards */
2892 *--retval = '\0';
2893 *--retval = ')';
2894 do {
2895 *--retval = PL_hexdigit[addr & 15];
2896 } while (addr >>= 4);
2897 *--retval = 'x';
2898 *--retval = '0';
2899 *--retval = '(';
2900
2901 retval -= typelen;
2902 memcpy(retval, typestr, typelen);
2903
2904 if (stashname) {
2905 *--retval = '=';
2906 retval -= stashnamelen;
2907 memcpy(retval, stashname, stashnamelen);
2908 }
2909 /* retval may not neccesarily have reached the start of the
2910 buffer here. */
2911 assert (retval >= buffer);
2912
2913 len = buffer_end - retval - 1; /* -1 for that \0 */
c080367d 2914 }
042dae7a 2915 if (lp)
fafee734
NC
2916 *lp = len;
2917 SAVEFREEPV(buffer);
2918 return retval;
463ee0b2 2919 }
79072805 2920 }
0336b60e 2921 if (SvREADONLY(sv) && !SvOK(sv)) {
cdb061a3
NC
2922 if (lp)
2923 *lp = 0;
9f621bb0
NC
2924 if (flags & SV_UNDEF_RETURNS_NULL)
2925 return NULL;
2926 if (ckWARN(WARN_UNINITIALIZED))
2927 report_uninit(sv);
73d840c0 2928 return (char *)"";
79072805 2929 }
79072805 2930 }
28e5dec8
JH
2931 if (SvIOK(sv) || ((SvIOKp(sv) && !SvNOKp(sv)))) {
2932 /* I'm assuming that if both IV and NV are equally valid then
2933 converting the IV is going to be more efficient */
e1ec3a88 2934 const U32 isUIOK = SvIsUV(sv);
28e5dec8
JH
2935 char buf[TYPE_CHARS(UV)];
2936 char *ebuf, *ptr;
97a130b8 2937 STRLEN len;
28e5dec8
JH
2938
2939 if (SvTYPE(sv) < SVt_PVIV)
2940 sv_upgrade(sv, SVt_PVIV);
4ea1d550 2941 ptr = uiv_2buf(buf, SvIVX(sv), SvUVX(sv), isUIOK, &ebuf);
97a130b8 2942 len = ebuf - ptr;
5902b6a9 2943 /* inlined from sv_setpvn */
97a130b8
NC
2944 s = SvGROW_mutable(sv, len + 1);
2945 Move(ptr, s, len, char);
2946 s += len;
28e5dec8 2947 *s = '\0';
28e5dec8
JH
2948 }
2949 else if (SvNOKp(sv)) {
4ee39169 2950 dSAVE_ERRNO;
79072805
LW
2951 if (SvTYPE(sv) < SVt_PVNV)
2952 sv_upgrade(sv, SVt_PVNV);
1c846c1f 2953 /* The +20 is pure guesswork. Configure test needed. --jhi */
5902b6a9 2954 s = SvGROW_mutable(sv, NV_DIG + 20);
c81271c3 2955 /* some Xenix systems wipe out errno here */
79072805 2956#ifdef apollo
463ee0b2 2957 if (SvNVX(sv) == 0.0)
d1307786 2958 my_strlcpy(s, "0", SvLEN(sv));
79072805
LW
2959 else
2960#endif /*apollo*/
bbce6d69 2961 {
2d4389e4 2962 Gconvert(SvNVX(sv), NV_DIG, 0, s);
bbce6d69 2963 }
4ee39169 2964 RESTORE_ERRNO;
a0d0e21e 2965#ifdef FIXNEGATIVEZERO
20773dcd
NC
2966 if (*s == '-' && s[1] == '0' && !s[2]) {
2967 s[0] = '0';
2968 s[1] = 0;
2969 }
a0d0e21e 2970#endif
79072805
LW
2971 while (*s) s++;
2972#ifdef hcx
2973 if (s[-1] == '.')
46fc3d4c 2974 *--s = '\0';
79072805
LW
2975#endif
2976 }
79072805 2977 else {
8d1c3e26
NC
2978 if (isGV_with_GP(sv)) {
2979 GV *const gv = MUTABLE_GV(sv);
2980 const U32 wasfake = SvFLAGS(gv) & SVf_FAKE;
2981 SV *const buffer = sv_newmortal();
2982
2983 /* FAKE globs can get coerced, so need to turn this off temporarily
2984 if it is on. */
2985 SvFAKE_off(gv);
2986 gv_efullname3(buffer, gv, "*");
2987 SvFLAGS(gv) |= wasfake;
2988
2989 assert(SvPOK(buffer));
2990 if (lp) {
2991 *lp = SvCUR(buffer);
2992 }
2993 return SvPVX(buffer);
2994 }
180488f8 2995
cdb061a3 2996 if (lp)
00b6aa41 2997 *lp = 0;
9f621bb0
NC
2998 if (flags & SV_UNDEF_RETURNS_NULL)
2999 return NULL;
3000 if (!PL_localizing && !(SvFLAGS(sv) & SVs_PADTMP) && ckWARN(WARN_UNINITIALIZED))
3001 report_uninit(sv);
25da4f38
IZ
3002 if (SvTYPE(sv) < SVt_PV)
3003 /* Typically the caller expects that sv_any is not NULL now. */
3004 sv_upgrade(sv, SVt_PV);
73d840c0 3005 return (char *)"";
79072805 3006 }
cdb061a3 3007 {
823a54a3 3008 const STRLEN len = s - SvPVX_const(sv);
cdb061a3
NC
3009 if (lp)
3010 *lp = len;
3011 SvCUR_set(sv, len);
3012 }
79072805 3013 SvPOK_on(sv);
1d7c1841 3014 DEBUG_c(PerlIO_printf(Perl_debug_log, "0x%"UVxf" 2pv(%s)\n",
3f7c398e 3015 PTR2UV(sv),SvPVX_const(sv)));
4d84ee25
NC
3016 if (flags & SV_CONST_RETURN)
3017 return (char *)SvPVX_const(sv);
10516c54
NC
3018 if (flags & SV_MUTABLE_RETURN)
3019 return SvPVX_mutable(sv);
463ee0b2
LW
3020 return SvPVX(sv);
3021}
3022
645c22ef 3023/*
6050d10e
JP
3024=for apidoc sv_copypv
3025
3026Copies a stringified representation of the source SV into the
3027destination SV. Automatically performs any necessary mg_get and
54f0641b 3028coercion of numeric values into strings. Guaranteed to preserve
2575c402 3029UTF8 flag even from overloaded objects. Similar in nature to
54f0641b
NIS
3030sv_2pv[_flags] but operates directly on an SV instead of just the
3031string. Mostly uses sv_2pv_flags to do its work, except when that
6050d10e
JP
3032would lose the UTF-8'ness of the PV.
3033
3034=cut
3035*/
3036
3037void
5de3775c 3038Perl_sv_copypv(pTHX_ SV *const dsv, register SV *const ssv)
6050d10e 3039{
446eaa42 3040 STRLEN len;
53c1dcc0 3041 const char * const s = SvPV_const(ssv,len);
7918f24d
NC
3042
3043 PERL_ARGS_ASSERT_SV_COPYPV;
3044
cb50f42d 3045 sv_setpvn(dsv,s,len);
446eaa42 3046 if (SvUTF8(ssv))
cb50f42d 3047 SvUTF8_on(dsv);
446eaa42 3048 else
cb50f42d 3049 SvUTF8_off(dsv);
6050d10e
JP
3050}
3051
3052/*
645c22ef
DM
3053=for apidoc sv_2pvbyte
3054
3055Return a pointer to the byte-encoded representation of the SV, and set *lp
1e54db1a 3056to its length. May cause the SV to be downgraded from UTF-8 as a
645c22ef
DM
3057side-effect.
3058
3059Usually accessed via the C<SvPVbyte> macro.
3060
3061=cut
3062*/
3063
7340a771 3064char *
5de3775c 3065Perl_sv_2pvbyte(pTHX_ register SV *const sv, STRLEN *const lp)
7340a771 3066{
7918f24d
NC
3067 PERL_ARGS_ASSERT_SV_2PVBYTE;
3068
0875d2fe 3069 sv_utf8_downgrade(sv,0);
97972285 3070 return lp ? SvPV(sv,*lp) : SvPV_nolen(sv);
7340a771
GS
3071}
3072
645c22ef 3073/*
035cbb0e
RGS
3074=for apidoc sv_2pvutf8
3075
3076Return a pointer to the UTF-8-encoded representation of the SV, and set *lp
3077to its length. May cause the SV to be upgraded to UTF-8 as a side-effect.
3078
3079Usually accessed via the C<SvPVutf8> macro.
3080
3081=cut
3082*/
645c22ef 3083
7340a771 3084char *
7bc54cea 3085Perl_sv_2pvutf8(pTHX_ register SV *const sv, STRLEN *const lp)
7340a771 3086{
7918f24d
NC
3087 PERL_ARGS_ASSERT_SV_2PVUTF8;
3088
035cbb0e
RGS
3089 sv_utf8_upgrade(sv);
3090 return lp ? SvPV(sv,*lp) : SvPV_nolen(sv);
7340a771 3091}
1c846c1f 3092
7ee2227d 3093
645c22ef
DM
3094/*
3095=for apidoc sv_2bool
3096
3097This function is only called on magical items, and is only used by
8cf8f3d1 3098sv_true() or its macro equivalent.
645c22ef
DM
3099
3100=cut
3101*/
3102
463ee0b2 3103bool
7bc54cea 3104Perl_sv_2bool(pTHX_ register SV *const sv)
463ee0b2 3105{
97aff369 3106 dVAR;
7918f24d
NC
3107
3108 PERL_ARGS_ASSERT_SV_2BOOL;
3109
5b295bef 3110 SvGETMAGIC(sv);
463ee0b2 3111
a0d0e21e
LW
3112 if (!SvOK(sv))
3113 return 0;
3114 if (SvROK(sv)) {
fabdb6c0
AL
3115 if (SvAMAGIC(sv)) {
3116 SV * const tmpsv = AMG_CALLun(sv,bool_);
3117 if (tmpsv && (!SvROK(tmpsv) || (SvRV(tmpsv) != SvRV(sv))))
3118 return (bool)SvTRUE(tmpsv);
3119 }
3120 return SvRV(sv) != 0;
a0d0e21e 3121 }
463ee0b2 3122 if (SvPOKp(sv)) {
53c1dcc0
AL
3123 register XPV* const Xpvtmp = (XPV*)SvANY(sv);
3124 if (Xpvtmp &&
339049b0 3125 (*sv->sv_u.svu_pv > '0' ||
11343788 3126 Xpvtmp->xpv_cur > 1 ||
339049b0 3127 (Xpvtmp->xpv_cur && *sv->sv_u.svu_pv != '0')))
463ee0b2
LW
3128 return 1;
3129 else
3130 return 0;
3131 }
3132 else {
3133 if (SvIOKp(sv))
3134 return SvIVX(sv) != 0;
3135 else {
3136 if (SvNOKp(sv))
3137 return SvNVX(sv) != 0.0;
180488f8 3138 else {
f7877b28 3139 if (isGV_with_GP(sv))
180488f8
NC
3140 return TRUE;
3141 else
3142 return FALSE;
3143 }
463ee0b2
LW
3144 }
3145 }
79072805
LW
3146}
3147
c461cf8f
JH
3148/*
3149=for apidoc sv_utf8_upgrade
3150
78ea37eb 3151Converts the PV of an SV to its UTF-8-encoded form.
645c22ef 3152Forces the SV to string form if it is not already.
2bbc8d55 3153Will C<mg_get> on C<sv> if appropriate.
4411f3b6 3154Always sets the SvUTF8 flag to avoid future validity checks even
2bbc8d55
SP
3155if the whole string is the same in UTF-8 as not.
3156Returns the number of bytes in the converted string
c461cf8f 3157
13a6c0e0
JH
3158This is not as a general purpose byte encoding to Unicode interface:
3159use the Encode extension for that.
3160
fe749c9a
KW
3161=for apidoc sv_utf8_upgrade_nomg
3162
3163Like sv_utf8_upgrade, but doesn't do magic on C<sv>
3164
8d6d96c1
HS
3165=for apidoc sv_utf8_upgrade_flags
3166
78ea37eb 3167Converts the PV of an SV to its UTF-8-encoded form.
645c22ef 3168Forces the SV to string form if it is not already.
8d6d96c1 3169Always sets the SvUTF8 flag to avoid future validity checks even
2bbc8d55
SP
3170if all the bytes are invariant in UTF-8. If C<flags> has C<SV_GMAGIC> bit set,
3171will C<mg_get> on C<sv> if appropriate, else not.
3172Returns the number of bytes in the converted string
3173C<sv_utf8_upgrade> and
8d6d96c1
HS
3174C<sv_utf8_upgrade_nomg> are implemented in terms of this function.
3175
13a6c0e0
JH
3176This is not as a general purpose byte encoding to Unicode interface:
3177use the Encode extension for that.
3178
8d6d96c1 3179=cut
b3ab6785
KW
3180
3181The grow version is currently not externally documented. It adds a parameter,
3182extra, which is the number of unused bytes the string of 'sv' is guaranteed to
3183have free after it upon return. This allows the caller to reserve extra space
3184that it intends to fill, to avoid extra grows.
3185
3186Also externally undocumented for the moment is the flag SV_FORCE_UTF8_UPGRADE,
3187which can be used to tell this function to not first check to see if there are
3188any characters that are different in UTF-8 (variant characters) which would
3189force it to allocate a new string to sv, but to assume there are. Typically
3190this flag is used by a routine that has already parsed the string to find that
3191there are such characters, and passes this information on so that the work
3192doesn't have to be repeated.
3193
3194(One might think that the calling routine could pass in the position of the
3195first such variant, so it wouldn't have to be found again. But that is not the
3196case, because typically when the caller is likely to use this flag, it won't be
3197calling this routine unless it finds something that won't fit into a byte.
3198Otherwise it tries to not upgrade and just use bytes. But some things that
3199do fit into a byte are variants in utf8, and the caller may not have been
3200keeping track of these.)
3201
3202If the routine itself changes the string, it adds a trailing NUL. Such a NUL
3203isn't guaranteed due to having other routines do the work in some input cases,
3204or if the input is already flagged as being in utf8.
3205
3206The speed of this could perhaps be improved for many cases if someone wanted to
3207write a fast function that counts the number of variant characters in a string,
3208especially if it could return the position of the first one.
3209
8d6d96c1
HS
3210*/
3211
3212STRLEN
b3ab6785 3213Perl_sv_utf8_upgrade_flags_grow(pTHX_ register SV *const sv, const I32 flags, STRLEN extra)
8d6d96c1 3214{
97aff369 3215 dVAR;
7918f24d 3216
b3ab6785 3217 PERL_ARGS_ASSERT_SV_UTF8_UPGRADE_FLAGS_GROW;
7918f24d 3218
808c356f
RGS
3219 if (sv == &PL_sv_undef)
3220 return 0;
e0e62c2a
NIS
3221 if (!SvPOK(sv)) {
3222 STRLEN len = 0;
d52b7888
NC
3223 if (SvREADONLY(sv) && (SvPOKp(sv) || SvIOKp(sv) || SvNOKp(sv))) {
3224 (void) sv_2pv_flags(sv,&len, flags);
b3ab6785
KW
3225 if (SvUTF8(sv)) {
3226 if (extra) SvGROW(sv, SvCUR(sv) + extra);
d52b7888 3227 return len;
b3ab6785 3228 }
d52b7888
NC
3229 } else {
3230 (void) SvPV_force(sv,len);
3231 }
e0e62c2a 3232 }
4411f3b6 3233
f5cee72b 3234 if (SvUTF8(sv)) {
b3ab6785 3235 if (extra) SvGROW(sv, SvCUR(sv) + extra);
5fec3b1d 3236 return SvCUR(sv);
f5cee72b 3237 }
5fec3b1d 3238
765f542d
NC
3239 if (SvIsCOW(sv)) {
3240 sv_force_normal_flags(sv, 0);
db42d148
NIS
3241 }
3242
b3ab6785 3243 if (PL_encoding && !(flags & SV_UTF8_NO_ENCODING)) {
799ef3cb 3244 sv_recode_to_utf8(sv, PL_encoding);
b3ab6785
KW
3245 if (extra) SvGROW(sv, SvCUR(sv) + extra);
3246 return SvCUR(sv);
3247 }
3248
4e93345f
KW
3249 if (SvCUR(sv) == 0) {
3250 if (extra) SvGROW(sv, extra);
3251 } else { /* Assume Latin-1/EBCDIC */
c4e7c712 3252 /* This function could be much more efficient if we
2bbc8d55 3253 * had a FLAG in SVs to signal if there are any variant
c4e7c712 3254 * chars in the PV. Given that there isn't such a flag
b3ab6785
KW
3255 * make the loop as fast as possible (although there are certainly ways
3256 * to speed this up, eg. through vectorization) */
3257 U8 * s = (U8 *) SvPVX_const(sv);
3258 U8 * e = (U8 *) SvEND(sv);
3259 U8 *t = s;
3260 STRLEN two_byte_count = 0;
c4e7c712 3261
b3ab6785
KW
3262 if (flags & SV_FORCE_UTF8_UPGRADE) goto must_be_utf8;
3263
3264 /* See if really will need to convert to utf8. We mustn't rely on our
3265 * incoming SV being well formed and having a trailing '\0', as certain
3266 * code in pp_formline can send us partially built SVs. */
3267
c4e7c712 3268 while (t < e) {
53c1dcc0 3269 const U8 ch = *t++;
b3ab6785
KW
3270 if (NATIVE_IS_INVARIANT(ch)) continue;
3271
3272 t--; /* t already incremented; re-point to first variant */
3273 two_byte_count = 1;
3274 goto must_be_utf8;
c4e7c712 3275 }
b3ab6785
KW
3276
3277 /* utf8 conversion not needed because all are invariants. Mark as
3278 * UTF-8 even if no variant - saves scanning loop */
c4e7c712 3279 SvUTF8_on(sv);
b3ab6785
KW
3280 return SvCUR(sv);
3281
3282must_be_utf8:
3283
3284 /* Here, the string should be converted to utf8, either because of an
3285 * input flag (two_byte_count = 0), or because a character that
3286 * requires 2 bytes was found (two_byte_count = 1). t points either to
3287 * the beginning of the string (if we didn't examine anything), or to
3288 * the first variant. In either case, everything from s to t - 1 will
3289 * occupy only 1 byte each on output.
3290 *
3291 * There are two main ways to convert. One is to create a new string
3292 * and go through the input starting from the beginning, appending each
3293 * converted value onto the new string as we go along. It's probably
3294 * best to allocate enough space in the string for the worst possible
3295 * case rather than possibly running out of space and having to
3296 * reallocate and then copy what we've done so far. Since everything
3297 * from s to t - 1 is invariant, the destination can be initialized
3298 * with these using a fast memory copy
3299 *
3300 * The other way is to figure out exactly how big the string should be
3301 * by parsing the entire input. Then you don't have to make it big
3302 * enough to handle the worst possible case, and more importantly, if
3303 * the string you already have is large enough, you don't have to
3304 * allocate a new string, you can copy the last character in the input
3305 * string to the final position(s) that will be occupied by the
3306 * converted string and go backwards, stopping at t, since everything
3307 * before that is invariant.
3308 *
3309 * There are advantages and disadvantages to each method.
3310 *
3311 * In the first method, we can allocate a new string, do the memory
3312 * copy from the s to t - 1, and then proceed through the rest of the
3313 * string byte-by-byte.
3314 *
3315 * In the second method, we proceed through the rest of the input
3316 * string just calculating how big the converted string will be. Then
3317 * there are two cases:
3318 * 1) if the string has enough extra space to handle the converted
3319 * value. We go backwards through the string, converting until we
3320 * get to the position we are at now, and then stop. If this
3321 * position is far enough along in the string, this method is
3322 * faster than the other method. If the memory copy were the same
3323 * speed as the byte-by-byte loop, that position would be about
3324 * half-way, as at the half-way mark, parsing to the end and back
3325 * is one complete string's parse, the same amount as starting
3326 * over and going all the way through. Actually, it would be
3327 * somewhat less than half-way, as it's faster to just count bytes
3328 * than to also copy, and we don't have the overhead of allocating
3329 * a new string, changing the scalar to use it, and freeing the
3330 * existing one. But if the memory copy is fast, the break-even
3331 * point is somewhere after half way. The counting loop could be
3332 * sped up by vectorization, etc, to move the break-even point
3333 * further towards the beginning.
3334 * 2) if the string doesn't have enough space to handle the converted
3335 * value. A new string will have to be allocated, and one might
3336 * as well, given that, start from the beginning doing the first
3337 * method. We've spent extra time parsing the string and in
3338 * exchange all we've gotten is that we know precisely how big to
3339 * make the new one. Perl is more optimized for time than space,
3340 * so this case is a loser.
3341 * So what I've decided to do is not use the 2nd method unless it is
3342 * guaranteed that a new string won't have to be allocated, assuming
3343 * the worst case. I also decided not to put any more conditions on it
3344 * than this, for now. It seems likely that, since the worst case is
3345 * twice as big as the unknown portion of the string (plus 1), we won't
3346 * be guaranteed enough space, causing us to go to the first method,
3347 * unless the string is short, or the first variant character is near
3348 * the end of it. In either of these cases, it seems best to use the
3349 * 2nd method. The only circumstance I can think of where this would
3350 * be really slower is if the string had once had much more data in it
3351 * than it does now, but there is still a substantial amount in it */
3352
3353 {
3354 STRLEN invariant_head = t - s;
3355 STRLEN size = invariant_head + (e - t) * 2 + 1 + extra;
3356 if (SvLEN(sv) < size) {
3357
3358 /* Here, have decided to allocate a new string */
3359
3360 U8 *dst;
3361 U8 *d;
3362
3363 Newx(dst, size, U8);
3364
3365 /* If no known invariants at the beginning of the input string,
3366 * set so starts from there. Otherwise, can use memory copy to
3367 * get up to where we are now, and then start from here */
3368
3369 if (invariant_head <= 0) {
3370 d = dst;
3371 } else {
3372 Copy(s, dst, invariant_head, char);
3373 d = dst + invariant_head;
3374 }
3375
3376 while (t < e) {
3377 const UV uv = NATIVE8_TO_UNI(*t++);
3378 if (UNI_IS_INVARIANT(uv))
3379 *d++ = (U8)UNI_TO_NATIVE(uv);
3380 else {
3381 *d++ = (U8)UTF8_EIGHT_BIT_HI(uv);
3382 *d++ = (U8)UTF8_EIGHT_BIT_LO(uv);
3383 }
3384 }
3385 *d = '\0';
3386 SvPV_free(sv); /* No longer using pre-existing string */
3387 SvPV_set(sv, (char*)dst);
3388 SvCUR_set(sv, d - dst);
3389 SvLEN_set(sv, size);
3390 } else {
3391
3392 /* Here, have decided to get the exact size of the string.
3393 * Currently this happens only when we know that there is
3394 * guaranteed enough space to fit the converted string, so
3395 * don't have to worry about growing. If two_byte_count is 0,
3396 * then t points to the first byte of the string which hasn't
3397 * been examined yet. Otherwise two_byte_count is 1, and t
3398 * points to the first byte in the string that will expand to
3399 * two. Depending on this, start examining at t or 1 after t.
3400 * */
3401
3402 U8 *d = t + two_byte_count;
3403
3404
3405 /* Count up the remaining bytes that expand to two */
3406
3407 while (d < e) {
3408 const U8 chr = *d++;
3409 if (! NATIVE_IS_INVARIANT(chr)) two_byte_count++;
3410 }
3411
3412 /* The string will expand by just the number of bytes that
3413 * occupy two positions. But we are one afterwards because of
3414 * the increment just above. This is the place to put the
3415 * trailing NUL, and to set the length before we decrement */
3416
3417 d += two_byte_count;
3418 SvCUR_set(sv, d - s);
3419 *d-- = '\0';
3420
3421
3422 /* Having decremented d, it points to the position to put the
3423 * very last byte of the expanded string. Go backwards through
3424 * the string, copying and expanding as we go, stopping when we
3425 * get to the part that is invariant the rest of the way down */
3426
3427 e--;
3428 while (e >= t) {
3429 const U8 ch = NATIVE8_TO_UNI(*e--);
3430 if (UNI_IS_INVARIANT(ch)) {
3431 *d-- = UNI_TO_NATIVE(ch);
3432 } else {
3433 *d-- = (U8)UTF8_EIGHT_BIT_LO(ch);
3434 *d-- = (U8)UTF8_EIGHT_BIT_HI(ch);
3435 }
3436 }
3437 }
3438 }
560a288e 3439 }
b3ab6785
KW
3440
3441 /* Mark as UTF-8 even if no variant - saves scanning loop */
3442 SvUTF8_on(sv);
4411f3b6 3443 return SvCUR(sv);
560a288e
GS
3444}
3445
c461cf8f
JH
3446/*
3447=for apidoc sv_utf8_downgrade
3448
78ea37eb 3449Attempts to convert the PV of an SV from characters to bytes.
2bbc8d55
SP
3450If the PV contains a character that cannot fit
3451in a byte, this conversion will fail;
78ea37eb 3452in this case, either returns false or, if C<fail_ok> is not
c461cf8f
JH
3453true, croaks.
3454
13a6c0e0
JH
3455This is not as a general purpose Unicode to byte encoding interface:
3456use the Encode extension for that.
3457
c461cf8f
JH
3458=cut
3459*/
3460
560a288e 3461bool
7bc54cea 3462Perl_sv_utf8_downgrade(pTHX_ register SV *const sv, const bool fail_ok)
560a288e 3463{
97aff369 3464 dVAR;
7918f24d
NC
3465
3466 PERL_ARGS_ASSERT_SV_UTF8_DOWNGRADE;
3467
78ea37eb 3468 if (SvPOKp(sv) && SvUTF8(sv)) {
fa301091 3469 if (SvCUR(sv)) {
03cfe0ae 3470 U8 *s;
652088fc 3471 STRLEN len;
fa301091 3472
765f542d
NC
3473 if (SvIsCOW(sv)) {
3474 sv_force_normal_flags(sv, 0);
3475 }
03cfe0ae
NIS
3476 s = (U8 *) SvPV(sv, len);
3477 if (!utf8_to_bytes(s, &len)) {
fa301091
JH
3478 if (fail_ok)
3479 return FALSE;
3480 else {
3481 if (PL_op)
3482 Perl_croak(aTHX_ "Wide character in %s",
53e06cf0 3483 OP_DESC(PL_op));
fa301091
JH
3484 else
3485 Perl_croak(aTHX_ "Wide character");
3486 }
4b3603a4 3487 }
b162af07 3488 SvCUR_set(sv, len);
67e989fb 3489 }
560a288e 3490 }
ffebcc3e 3491 SvUTF8_off(sv);
560a288e
GS
3492 return TRUE;
3493}
3494
c461cf8f
JH
3495/*
3496=for apidoc sv_utf8_encode
3497
78ea37eb
TS
3498Converts the PV of an SV to UTF-8, but then turns the C<SvUTF8>
3499flag off so that it looks like octets again.
c461cf8f
JH
3500
3501=cut
3502*/
3503
560a288e 3504void
7bc54cea 3505Perl_sv_utf8_encode(pTHX_ register SV *const sv)
560a288e 3506{
7918f24d
NC
3507 PERL_ARGS_ASSERT_SV_UTF8_ENCODE;
3508
4c94c214
NC
3509 if (SvIsCOW(sv)) {
3510 sv_force_normal_flags(sv, 0);
3511 }
3512 if (SvREADONLY(sv)) {
f1f66076 3513 Perl_croak(aTHX_ "%s", PL_no_modify);
4c94c214 3514 }
a5f5288a 3515 (void) sv_utf8_upgrade(sv);
560a288e
GS
3516 SvUTF8_off(sv);
3517}
3518
4411f3b6
NIS
3519/*
3520=for apidoc sv_utf8_decode
3521
78ea37eb
TS
3522If the PV of the SV is an octet sequence in UTF-8
3523and contains a multiple-byte character, the C<SvUTF8> flag is turned on
3524so that it looks like a character. If the PV contains only single-byte
3525characters, the C<SvUTF8> flag stays being off.
3526Scans PV for validity and returns false if the PV is invalid UTF-8.
4411f3b6
NIS
3527
3528=cut
3529*/
3530
560a288e 3531bool
7bc54cea 3532Perl_sv_utf8_decode(pTHX_ register SV *const sv)
560a288e 3533{
7918f24d
NC
3534 PERL_ARGS_ASSERT_SV_UTF8_DECODE;
3535
78ea37eb 3536 if (SvPOKp(sv)) {
93524f2b
NC
3537 const U8 *c;
3538 const U8 *e;
9cbac4c7 3539
645c22ef
DM
3540 /* The octets may have got themselves encoded - get them back as
3541 * bytes
3542 */
3543 if (!sv_utf8_downgrade(sv, TRUE))
560a288e
GS
3544 return FALSE;
3545
3546 /* it is actually just a matter of turning the utf8 flag on, but
3547 * we want to make sure everything inside is valid utf8 first.
3548 */
93524f2b 3549 c = (const U8 *) SvPVX_const(sv);
63cd0674 3550 if (!is_utf8_string(c, SvCUR(sv)+1))
67e989fb 3551 return FALSE;
93524f2b 3552 e = (const U8 *) SvEND(sv);
511c2ff0 3553 while (c < e) {
b64e5050 3554 const U8 ch = *c++;
c4d5f83a 3555 if (!UTF8_IS_INVARIANT(ch)) {
67e989fb
JH
3556 SvUTF8_on(sv);
3557 break;
3558 }
560a288e 3559 }
560a288e
GS
3560 }
3561 return TRUE;
3562}
3563
954c1994
GS
3564/*
3565=for apidoc sv_setsv
3566
645c22ef
DM
3567Copies the contents of the source SV C<ssv> into the destination SV
3568C<dsv>. The source SV may be destroyed if it is mortal, so don't use this
3569function if the source SV needs to be reused. Does not handle 'set' magic.
3570Loosely speaking, it performs a copy-by-value, obliterating any previous
3571content of the destination.
3572
3573You probably want to use one of the assortment of wrappers, such as
3574C<SvSetSV>, C<SvSetSV_nosteal>, C<SvSetMagicSV> and
3575C<SvSetMagicSV_nosteal>.
3576
8d6d96c1
HS
3577=for apidoc sv_setsv_flags
3578
645c22ef
DM
3579Copies the contents of the source SV C<ssv> into the destination SV
3580C<dsv>. The source SV may be destroyed if it is mortal, so don't use this
3581function if the source SV needs to be reused. Does not handle 'set' magic.
3582Loosely speaking, it performs a copy-by-value, obliterating any previous
3583content of the destination.
3584If the C<flags> parameter has the C<SV_GMAGIC> bit set, will C<mg_get> on
5fcdf167
NC
3585C<ssv> if appropriate, else not. If the C<flags> parameter has the
3586C<NOSTEAL> bit set then the buffers of temps will not be stolen. <sv_setsv>
3587and C<sv_setsv_nomg> are implemented in terms of this function.
645c22ef
DM
3588
3589You probably want to use one of the assortment of wrappers, such as
3590C<SvSetSV>, C<SvSetSV_nosteal>, C<SvSetMagicSV> and
3591C<SvSetMagicSV_nosteal>.
3592
3593This is the primary function for copying scalars, and most other
3594copy-ish functions and macros use this underneath.
8d6d96c1
HS
3595
3596=cut
3597*/
3598
5d0301b7 3599static void
7bc54cea 3600S_glob_assign_glob(pTHX_ SV *const dstr, SV *const sstr, const int dtype)
5d0301b7 3601{
70cd14a1 3602 I32 mro_changes = 0; /* 1 = method, 2 = isa */
dd69841b 3603
7918f24d
NC
3604 PERL_ARGS_ASSERT_GLOB_ASSIGN_GLOB;
3605
5d0301b7
NC
3606 if (dtype != SVt_PVGV) {
3607 const char * const name = GvNAME(sstr);
3608 const STRLEN len = GvNAMELEN(sstr);
0d092c36 3609 {
f7877b28
NC
3610 if (dtype >= SVt_PV) {
3611 SvPV_free(dstr);
3612 SvPV_set(dstr, 0);
3613 SvLEN_set(dstr, 0);
3614 SvCUR_set(dstr, 0);
3615 }
0d092c36 3616 SvUPGRADE(dstr, SVt_PVGV);
dedf8e73 3617 (void)SvOK_off(dstr);
2e5b91de
NC
3618 /* FIXME - why are we doing this, then turning it off and on again
3619 below? */
3620 isGV_with_GP_on(dstr);
f7877b28 3621 }
5d0301b7
NC
3622 GvSTASH(dstr) = GvSTASH(sstr);
3623 if (GvSTASH(dstr))
daba3364 3624 Perl_sv_add_backref(aTHX_ MUTABLE_SV(GvSTASH(dstr)), dstr);
159b6efe 3625 gv_name_set(MUTABLE_GV(dstr), name, len, GV_ADD);
5d0301b7
NC
3626 SvFAKE_on(dstr); /* can coerce to non-glob */
3627 }
3628
159b6efe 3629 if(GvGP(MUTABLE_GV(sstr))) {
dd69841b
BB
3630 /* If source has method cache entry, clear it */
3631 if(GvCVGEN(sstr)) {
3632 SvREFCNT_dec(GvCV(sstr));
3633 GvCV(sstr) = NULL;
3634 GvCVGEN(sstr) = 0;
3635 }
3636 /* If source has a real method, then a method is
3637 going to change */
159b6efe 3638 else if(GvCV((const GV *)sstr)) {
70cd14a1 3639 mro_changes = 1;
dd69841b
BB
3640 }
3641 }
3642
3643 /* If dest already had a real method, that's a change as well */
159b6efe 3644 if(!mro_changes && GvGP(MUTABLE_GV(dstr)) && GvCVu((const GV *)dstr)) {
70cd14a1 3645 mro_changes = 1;
dd69841b
BB
3646 }
3647
159b6efe 3648 if(strEQ(GvNAME((const GV *)dstr),"ISA"))
70cd14a1
CB
3649 mro_changes = 2;
3650
159b6efe 3651 gp_free(MUTABLE_GV(dstr));
2e5b91de 3652 isGV_with_GP_off(dstr);
5d0301b7 3653 (void)SvOK_off(dstr);
2e5b91de 3654 isGV_with_GP_on(dstr);
dedf8e73 3655 GvINTRO_off(dstr); /* one-shot flag */
5d0301b7
NC
3656 GvGP(dstr) = gp_ref(GvGP(sstr));
3657 if (SvTAINTED(sstr))
3658 SvTAINT(dstr);
3659 if (GvIMPORTED(dstr) != GVf_IMPORTED
3660 && CopSTASH_ne(PL_curcop, GvSTASH(dstr)))
3661 {
3662 GvIMPORTED_on(dstr);
3663 }
3664 GvMULTI_on(dstr);
70cd14a1
CB
3665 if(mro_changes == 2) mro_isa_changed_in(GvSTASH(dstr));
3666 else if(mro_changes) mro_method_changed_in(GvSTASH(dstr));
5d0301b7
NC
3667 return;
3668}
3669
b8473700 3670static void
7bc54cea 3671S_glob_assign_ref(pTHX_ SV *const dstr, SV *const sstr)
7918f24d 3672{
b8473700
NC
3673 SV * const sref = SvREFCNT_inc(SvRV(sstr));
3674 SV *dref = NULL;
3675 const int intro = GvINTRO(dstr);
2440974c 3676 SV **location;
3386d083 3677 U8 import_flag = 0;
27242d61 3678 const U32 stype = SvTYPE(sref);
26d68d86 3679 bool mro_changes = FALSE;
27242d61 3680
7918f24d 3681 PERL_ARGS_ASSERT_GLOB_ASSIGN_REF;
b8473700 3682
b8473700
NC
3683 if (intro) {
3684 GvINTRO_off(dstr); /* one-shot flag */
3685 GvLINE(dstr) = CopLINE(PL_curcop);
159b6efe 3686 GvEGV(dstr) = MUTABLE_GV(dstr);
b8473700
NC
3687 }
3688 GvMULTI_on(dstr);
27242d61 3689 switch (stype) {
b8473700 3690 case SVt_PVCV:
27242d61
NC
3691 location = (SV **) &GvCV(dstr);
3692 import_flag = GVf_IMPORTED_CV;
3693 goto common;
3694 case SVt_PVHV:
3695 location = (SV **) &GvHV(dstr);
3696 import_flag = GVf_IMPORTED_HV;
3697 goto common;
3698 case SVt_PVAV:
3699 location = (SV **) &GvAV(dstr);
26d68d86
TC
3700 if (strEQ(GvNAME((GV*)dstr), "ISA"))
3701 mro_changes = TRUE;
27242d61
NC
3702 import_flag = GVf_IMPORTED_AV;
3703 goto common;
3704 case SVt_PVIO:
3705 location = (SV **) &GvIOp(dstr);
3706 goto common;
3707 case SVt_PVFM:
3708 location = (SV **) &GvFORM(dstr);
ef595a33 3709 goto common;
27242d61
NC
3710 default:
3711 location = &GvSV(dstr);
3712 import_flag = GVf_IMPORTED_SV;
3713 common:
b8473700 3714 if (intro) {
27242d61 3715 if (stype == SVt_PVCV) {
ea726b52 3716 /*if (GvCVGEN(dstr) && (GvCV(dstr) != (const CV *)sref || GvCVGEN(dstr))) {*/
5f2fca8a 3717 if (GvCVGEN(dstr)) {
27242d61
NC
3718 SvREFCNT_dec(GvCV(dstr));
3719 GvCV(dstr) = NULL;
3720 GvCVGEN(dstr) = 0; /* Switch off cacheness. */
27242d61 3721 }
b8473700 3722 }
27242d61 3723 SAVEGENERICSV(*location);
b8473700
NC
3724 }
3725 else
27242d61 3726 dref = *location;
5f2fca8a 3727 if (stype == SVt_PVCV && (*location != sref || GvCVGEN(dstr))) {
ea726b52 3728 CV* const cv = MUTABLE_CV(*location);
b8473700 3729 if (cv) {
159b6efe 3730 if (!GvCVGEN((const GV *)dstr) &&
b8473700
NC
3731 (CvROOT(cv) || CvXSUB(cv)))
3732 {
3733 /* Redefining a sub - warning is mandatory if
3734 it was a const and its value changed. */
ea726b52 3735 if (CvCONST(cv) && CvCONST((const CV *)sref)
126f53f3
NC
3736 && cv_const_sv(cv)
3737 == cv_const_sv((const CV *)sref)) {
6f207bd3 3738 NOOP;
b8473700
NC
3739 /* They are 2 constant subroutines generated from
3740 the same constant. This probably means that
3741 they are really the "same" proxy subroutine
3742 instantiated in 2 places. Most likely this is
3743 when a constant is exported twice. Don't warn.
3744 */
3745 }
3746 else if (ckWARN(WARN_REDEFINE)
3747 || (CvCONST(cv)
ea726b52 3748 && (!CvCONST((const CV *)sref)
b8473700 3749 || sv_cmp(cv_const_sv(cv),
126f53f3
NC
3750 cv_const_sv((const CV *)
3751 sref))))) {
b8473700 3752 Perl_warner(aTHX_ packWARN(WARN_REDEFINE),
10edeb5d
JH
3753 (const char *)
3754 (CvCONST(cv)
3755 ? "Constant subroutine %s::%s redefined"
3756 : "Subroutine %s::%s redefined"),
159b6efe
NC
3757 HvNAME_get(GvSTASH((const GV *)dstr)),
3758 GvENAME(MUTABLE_GV(dstr)));
b8473700
NC
3759 }
3760 }
3761 if (!intro)
159b6efe 3762 cv_ckproto_len(cv, (const GV *)dstr,
cbf82dd0
NC
3763 SvPOK(sref) ? SvPVX_const(sref) : NULL,
3764 SvPOK(sref) ? SvCUR(sref) : 0);
b8473700 3765 }
b8473700
NC
3766 GvCVGEN(dstr) = 0; /* Switch off cacheness. */
3767 GvASSUMECV_on(dstr);
dd69841b 3768 if(GvSTASH(dstr)) mro_method_changed_in(GvSTASH(dstr)); /* sub foo { 1 } sub bar { 2 } *bar = \&foo */
b8473700 3769 }
2440974c 3770 *location = sref;
3386d083
NC
3771 if (import_flag && !(GvFLAGS(dstr) & import_flag)
3772 && CopSTASH_ne(PL_curcop, GvSTASH(dstr))) {
3773 GvFLAGS(dstr) |= import_flag;
b8473700
NC
3774 }
3775 break;
3776 }
b37c2d43 3777 SvREFCNT_dec(dref);
b8473700
NC
3778 if (SvTAINTED(sstr))
3779 SvTAINT(dstr);
26d68d86 3780 if (mro_changes) mro_isa_changed_in(GvSTASH(dstr));
b8473700
NC
3781 return;
3782}
3783
8d6d96c1 3784void
7bc54cea 3785Perl_sv_setsv_flags(pTHX_ SV *dstr, register SV* sstr, const I32 flags)
8d6d96c1 3786{
97aff369 3787 dVAR;
8990e307
LW
3788 register U32 sflags;
3789 register int dtype;
42d0e0b7 3790 register svtype stype;
463ee0b2 3791
7918f24d
NC
3792 PERL_ARGS_ASSERT_SV_SETSV_FLAGS;
3793
79072805
LW
3794 if (sstr == dstr)
3795 return;
29f4f0ab
NC
3796
3797 if (SvIS_FREED(dstr)) {
3798 Perl_croak(aTHX_ "panic: attempt to copy value %" SVf
be2597df 3799 " to a freed scalar %p", SVfARG(sstr), (void *)dstr);
29f4f0ab 3800 }
765f542d 3801 SV_CHECK_THINKFIRST_COW_DROP(dstr);
79072805 3802 if (!sstr)
3280af22 3803 sstr = &PL_sv_undef;
29f4f0ab 3804 if (SvIS_FREED(sstr)) {
6c9570dc
MHM
3805 Perl_croak(aTHX_ "panic: attempt to copy freed scalar %p to %p",
3806 (void*)sstr, (void*)dstr);
29f4f0ab 3807 }
8990e307
LW
3808 stype = SvTYPE(sstr);
3809 dtype = SvTYPE(dstr);
79072805 3810
52944de8 3811 (void)SvAMAGIC_off(dstr);
7a5fa8a2 3812 if ( SvVOK(dstr) )
ece467f9
JP
3813 {
3814 /* need to nuke the magic */
3815 mg_free(dstr);
ece467f9 3816 }
9e7bc3e8 3817
463ee0b2 3818 /* There's a lot of redundancy below but we're going for speed here */
79072805 3819
8990e307 3820 switch (stype) {
79072805 3821 case SVt_NULL:
aece5585 3822 undef_sstr:
20408e3c
GS
3823 if (dtype != SVt_PVGV) {
3824 (void)SvOK_off(dstr);
3825 return;
3826 }
3827 break;
463ee0b2 3828 case SVt_IV:
aece5585
GA
3829 if (SvIOK(sstr)) {
3830 switch (dtype) {
3831 case SVt_NULL:
8990e307 3832 sv_upgrade(dstr, SVt_IV);
aece5585
GA
3833 break;
3834 case SVt_NV:
aece5585 3835 case SVt_PV:
a0d0e21e 3836 sv_upgrade(dstr, SVt_PVIV);
aece5585 3837 break;
010be86b
NC
3838 case SVt_PVGV:
3839 goto end_of_first_switch;
aece5585
GA
3840 }
3841 (void)SvIOK_only(dstr);
45977657 3842 SvIV_set(dstr, SvIVX(sstr));
25da4f38
IZ
3843 if (SvIsUV(sstr))
3844 SvIsUV_on(dstr);
37c25af0
NC
3845 /* SvTAINTED can only be true if the SV has taint magic, which in
3846 turn means that the SV type is PVMG (or greater). This is the
3847 case statement for SVt_IV, so this cannot be true (whatever gcov
3848 may say). */
3849 assert(!SvTAINTED(sstr));
aece5585 3850 return;
8990e307 3851 }
4df7f6af
NC
3852 if (!SvROK(sstr))
3853 goto undef_sstr;
3854 if (dtype < SVt_PV && dtype != SVt_IV)
3855 sv_upgrade(dstr, SVt_IV);
3856 break;
aece5585 3857
463ee0b2 3858 case SVt_NV:
aece5585
GA
3859 if (SvNOK(sstr)) {
3860 switch (dtype) {
3861 case SVt_NULL:
3862 case SVt_IV:
8990e307 3863 sv_upgrade(dstr, SVt_NV);
aece5585 3864 break;
aece5585
GA
3865 case SVt_PV:
3866 case SVt_PVIV:
a0d0e21e 3867 sv_upgrade(dstr, SVt_PVNV);
aece5585 3868 break;
010be86b
NC
3869 case SVt_PVGV:
3870 goto end_of_first_switch;
aece5585 3871 }
9d6ce603 3872 SvNV_set(dstr, SvNVX(sstr));
aece5585 3873 (void)SvNOK_only(dstr);
37c25af0
NC
3874 /* SvTAINTED can only be true if the SV has taint magic, which in
3875 turn means that the SV type is PVMG (or greater). This is the
3876 case statement for SVt_NV, so this cannot be true (whatever gcov
3877 may say). */
3878 assert(!SvTAINTED(sstr));
aece5585 3879 return;
8990e307 3880 }
aece5585
GA
3881 goto undef_sstr;
3882
fc36a67e 3883 case SVt_PVFM:
f8c7b90f 3884#ifdef PERL_OLD_COPY_ON_WRITE
d89fc664
NC
3885 if ((SvFLAGS(sstr) & CAN_COW_MASK) == CAN_COW_FLAGS) {
3886 if (dtype < SVt_PVIV)
3887 sv_upgrade(dstr, SVt_PVIV);
3888 break;
3889 }
3890 /* Fall through */
3891#endif
3892 case SVt_PV:
8990e307 3893 if (dtype < SVt_PV)
463ee0b2 3894 sv_upgrade(dstr, SVt_PV);
463ee0b2
LW
3895 break;
3896 case SVt_PVIV:
8990e307 3897 if (dtype < SVt_PVIV)
463ee0b2 3898 sv_upgrade(dstr, SVt_PVIV);
463ee0b2
LW
3899 break;
3900 case SVt_PVNV:
8990e307 3901 if (dtype < SVt_PVNV)
463ee0b2 3902 sv_upgrade(dstr, SVt_PVNV);
463ee0b2 3903 break;
489f7bfe 3904 default:
a3b680e6
AL
3905 {
3906 const char * const type = sv_reftype(sstr,0);
533c011a 3907 if (PL_op)
a3b680e6 3908 Perl_croak(aTHX_ "Bizarre copy of %s in %s", type, OP_NAME(PL_op));
4633a7c4 3909 else
a3b680e6
AL
3910 Perl_croak(aTHX_ "Bizarre copy of %s", type);
3911 }
4633a7c4
LW
3912 break;
3913
f0826785
BM
3914 case SVt_REGEXP:
3915 if (dtype < SVt_REGEXP)
3916 sv_upgrade(dstr, SVt_REGEXP);
3917 break;
3918
cecf5685 3919 /* case SVt_BIND: */
39cb70dc 3920 case SVt_PVLV:
79072805 3921 case SVt_PVGV:
cecf5685 3922 if (isGV_with_GP(sstr) && dtype <= SVt_PVGV) {
d4c19fe8 3923 glob_assign_glob(dstr, sstr, dtype);
b8c701c1 3924 return;
79072805 3925 }
cecf5685 3926 /* SvVALID means that this PVGV is playing at being an FBM. */
5f66b61c 3927 /*FALLTHROUGH*/
79072805 3928
489f7bfe 3929 case SVt_PVMG:
8d6d96c1 3930 if (SvGMAGICAL(sstr) && (flags & SV_GMAGIC)) {
973f89ab 3931 mg_get(sstr);
1d9c78c6 3932 if (SvTYPE(sstr) != stype) {
973f89ab 3933 stype = SvTYPE(sstr);
cecf5685 3934 if (isGV_with_GP(sstr) && stype == SVt_PVGV && dtype <= SVt_PVGV) {
d4c19fe8 3935 glob_assign_glob(dstr, sstr, dtype);
b8c701c1
NC
3936 return;
3937 }
973f89ab
CS
3938 }
3939 }
ded42b9f 3940 if (stype == SVt_PVLV)
862a34c6 3941 SvUPGRADE(dstr, SVt_PVNV);
ded42b9f 3942 else
42d0e0b7 3943 SvUPGRADE(dstr, (svtype)stype);
79072805 3944 }
010be86b 3945 end_of_first_switch:
79072805 3946
ff920335
NC
3947 /* dstr may have been upgraded. */
3948 dtype = SvTYPE(dstr);
8990e307
LW
3949 sflags = SvFLAGS(sstr);
3950
ba2fdce6 3951 if (dtype == SVt_PVCV || dtype == SVt_PVFM) {
85324b4d
NC
3952 /* Assigning to a subroutine sets the prototype. */
3953 if (SvOK(sstr)) {
3954 STRLEN len;
3955 const char *const ptr = SvPV_const(sstr, len);
3956
3957 SvGROW(dstr, len + 1);
3958 Copy(ptr, SvPVX(dstr), len + 1, char);
3959 SvCUR_set(dstr, len);
fcddd32e 3960 SvPOK_only(dstr);
ba2fdce6 3961 SvFLAGS(dstr) |= sflags & SVf_UTF8;
85324b4d
NC
3962 } else {
3963 SvOK_off(dstr);
3964 }
ba2fdce6
NC
3965 } else if (dtype == SVt_PVAV || dtype == SVt_PVHV) {
3966 const char * const type = sv_reftype(dstr,0);
3967 if (PL_op)
3968 Perl_croak(aTHX_ "Cannot copy to %s in %s", type, OP_NAME(PL_op));
3969 else
3970 Perl_croak(aTHX_ "Cannot copy to %s", type);
85324b4d 3971 } else if (sflags & SVf_ROK) {
cecf5685 3972 if (isGV_with_GP(dstr) && dtype == SVt_PVGV
785bee4f 3973 && SvTYPE(SvRV(sstr)) == SVt_PVGV && isGV_with_GP(SvRV(sstr))) {
acaa9288
NC
3974 sstr = SvRV(sstr);
3975 if (sstr == dstr) {
3976 if (GvIMPORTED(dstr) != GVf_IMPORTED
3977 && CopSTASH_ne(PL_curcop, GvSTASH(dstr)))
3978 {
3979 GvIMPORTED_on(dstr);
3980 }
3981 GvMULTI_on(dstr);
3982 return;
3983 }
785bee4f
NC
3984 glob_assign_glob(dstr, sstr, dtype);
3985 return;
acaa9288
NC
3986 }
3987
8990e307 3988 if (dtype >= SVt_PV) {
fdc5b023 3989 if (dtype == SVt_PVGV && isGV_with_GP(dstr)) {
d4c19fe8 3990 glob_assign_ref(dstr, sstr);
b8c701c1
NC
3991 return;
3992 }
3f7c398e 3993 if (SvPVX_const(dstr)) {
8bd4d4c5 3994 SvPV_free(dstr);
b162af07
SP
3995 SvLEN_set(dstr, 0);
3996 SvCUR_set(dstr, 0);
a0d0e21e 3997 }
8990e307 3998 }
a0d0e21e 3999 (void)SvOK_off(dstr);
b162af07 4000 SvRV_set(dstr, SvREFCNT_inc(SvRV(sstr)));
96d4b0ee 4001 SvFLAGS(dstr) |= sflags & SVf_ROK;
dfd48732
NC
4002 assert(!(sflags & SVp_NOK));
4003 assert(!(sflags & SVp_IOK));
4004 assert(!(sflags & SVf_NOK));
4005 assert(!(sflags & SVf_IOK));
ed6116ce 4006 }
cecf5685 4007 else if (dtype == SVt_PVGV && isGV_with_GP(dstr)) {
c0c44674 4008 if (!(sflags & SVf_OK)) {
a2a5de95
NC
4009 Perl_ck_warner(aTHX_ packWARN(WARN_MISC),
4010 "Undefined value assigned to typeglob");
c0c44674
NC
4011 }
4012 else {
4013 GV *gv = gv_fetchsv(sstr, GV_ADD, SVt_PVGV);
daba3364 4014 if (dstr != (const SV *)gv) {
c0c44674 4015 if (GvGP(dstr))
159b6efe 4016 gp_free(MUTABLE_GV(dstr));
c0c44674
NC
4017 GvGP(dstr) = gp_ref(GvGP(gv));
4018 }
4019 }
4020 }
f0826785
BM
4021 else if (dtype == SVt_REGEXP && stype == SVt_REGEXP) {
4022 reg_temp_copy((REGEXP*)dstr, (REGEXP*)sstr);
4023 }
8990e307 4024 else if (sflags & SVp_POK) {
765f542d 4025 bool isSwipe = 0;
79072805
LW
4026
4027 /*
4028 * Check to see if we can just swipe the string. If so, it's a
4029 * possible small lose on short strings, but a big win on long ones.
3f7c398e
SP
4030 * It might even be a win on short strings if SvPVX_const(dstr)
4031 * has to be allocated and SvPVX_const(sstr) has to be freed.
34482cd6
NC
4032 * Likewise if we can set up COW rather than doing an actual copy, we
4033 * drop to the else clause, as the swipe code and the COW setup code
4034 * have much in common.
79072805
LW
4035 */
4036
120fac95
NC
4037 /* Whichever path we take through the next code, we want this true,
4038 and doing it now facilitates the COW check. */
4039 (void)SvPOK_only(dstr);
4040
765f542d 4041 if (
34482cd6
NC
4042 /* If we're already COW then this clause is not true, and if COW
4043 is allowed then we drop down to the else and make dest COW
4044 with us. If caller hasn't said that we're allowed to COW
4045 shared hash keys then we don't do the COW setup, even if the
4046 source scalar is a shared hash key scalar. */
4047 (((flags & SV_COW_SHARED_HASH_KEYS)
4048 ? (sflags & (SVf_FAKE|SVf_READONLY)) != (SVf_FAKE|SVf_READONLY)
4049 : 1 /* If making a COW copy is forbidden then the behaviour we
4050 desire is as if the source SV isn't actually already
4051 COW, even if it is. So we act as if the source flags
4052 are not COW, rather than actually testing them. */
4053 )
f8c7b90f 4054#ifndef PERL_OLD_COPY_ON_WRITE
34482cd6
NC
4055 /* The change that added SV_COW_SHARED_HASH_KEYS makes the logic
4056 when PERL_OLD_COPY_ON_WRITE is defined a little wrong.
4057 Conceptually PERL_OLD_COPY_ON_WRITE being defined should
4058 override SV_COW_SHARED_HASH_KEYS, because it means "always COW"
4059 but in turn, it's somewhat dead code, never expected to go
4060 live, but more kept as a placeholder on how to do it better
4061 in a newer implementation. */
4062 /* If we are COW and dstr is a suitable target then we drop down
4063 into the else and make dest a COW of us. */
b8f9541a
NC
4064 || (SvFLAGS(dstr) & CAN_COW_MASK) != CAN_COW_FLAGS
4065#endif
4066 )
765f542d 4067 &&
765f542d
NC
4068 !(isSwipe =
4069 (sflags & SVs_TEMP) && /* slated for free anyway? */
4070 !(sflags & SVf_OOK) && /* and not involved in OOK hack? */
5fcdf167
NC
4071 (!(flags & SV_NOSTEAL)) &&
4072 /* and we're allowed to steal temps */
765f542d
NC
4073 SvREFCNT(sstr) == 1 && /* and no other references to it? */
4074 SvLEN(sstr) && /* and really is a string */
645c22ef 4075 /* and won't be needed again, potentially */
765f542d 4076 !(PL_op && PL_op->op_type == OP_AASSIGN))
f8c7b90f 4077#ifdef PERL_OLD_COPY_ON_WRITE
cb23d5b1
NC
4078 && ((flags & SV_COW_SHARED_HASH_KEYS)
4079 ? (!((sflags & CAN_COW_MASK) == CAN_COW_FLAGS
4080 && (SvFLAGS(dstr) & CAN_COW_MASK) == CAN_COW_FLAGS
4b1c7d9e 4081 && SvTYPE(sstr) >= SVt_PVIV && SvTYPE(sstr) != SVt_PVFM))
cb23d5b1 4082 : 1)
765f542d
NC
4083#endif
4084 ) {
4085 /* Failed the swipe test, and it's not a shared hash key either.
4086 Have to copy the string. */
4087 STRLEN len = SvCUR(sstr);
4088 SvGROW(dstr, len + 1); /* inlined from sv_setpvn */
3f7c398e 4089 Move(SvPVX_const(sstr),SvPVX(dstr),len,char);
765f542d
NC
4090 SvCUR_set(dstr, len);
4091 *SvEND(dstr) = '\0';
765f542d 4092 } else {
f8c7b90f 4093 /* If PERL_OLD_COPY_ON_WRITE is not defined, then isSwipe will always
765f542d 4094 be true in here. */
765f542d
NC
4095 /* Either it's a shared hash key, or it's suitable for
4096 copy-on-write or we can swipe the string. */
46187eeb 4097 if (DEBUG_C_TEST) {
ed252734 4098 PerlIO_printf(Perl_debug_log, "Copy on write: sstr --> dstr\n");
e419cbc5
NC
4099 sv_dump(sstr);
4100 sv_dump(dstr);
46187eeb 4101 }
f8c7b90f 4102#ifdef PERL_OLD_COPY_ON_WRITE
765f542d 4103 if (!isSwipe) {
765f542d
NC
4104 if ((sflags & (SVf_FAKE | SVf_READONLY))
4105 != (SVf_FAKE | SVf_READONLY)) {
4106 SvREADONLY_on(sstr);
4107 SvFAKE_on(sstr);
4108 /* Make the source SV into a loop of 1.
4109 (about to become 2) */
a29f6d03 4110 SV_COW_NEXT_SV_SET(sstr, sstr);
765f542d
NC
4111 }
4112 }
4113#endif
4114 /* Initial code is common. */
94010e71
NC
4115 if (SvPVX_const(dstr)) { /* we know that dtype >= SVt_PV */
4116 SvPV_free(dstr);
79072805 4117 }
765f542d 4118
765f542d
NC
4119 if (!isSwipe) {
4120 /* making another shared SV. */
4121 STRLEN cur = SvCUR(sstr);
4122 STRLEN len = SvLEN(sstr);
f8c7b90f 4123#ifdef PERL_OLD_COPY_ON_WRITE
765f542d 4124 if (len) {
b8f9541a 4125 assert (SvTYPE(dstr) >= SVt_PVIV);
765f542d
NC
4126 /* SvIsCOW_normal */
4127 /* splice us in between source and next-after-source. */
a29f6d03
NC
4128 SV_COW_NEXT_SV_SET(dstr, SV_COW_NEXT_SV(sstr));
4129 SV_COW_NEXT_SV_SET(sstr, dstr);
940132f3 4130 SvPV_set(dstr, SvPVX_mutable(sstr));
a604c751
NC
4131 } else
4132#endif
4133 {
765f542d 4134 /* SvIsCOW_shared_hash */
46187eeb
NC
4135 DEBUG_C(PerlIO_printf(Perl_debug_log,
4136 "Copy on write: Sharing hash\n"));
b8f9541a 4137
bdd68bc3 4138 assert (SvTYPE(dstr) >= SVt_PV);
765f542d 4139 SvPV_set(dstr,
d1db91c6 4140 HEK_KEY(share_hek_hek(SvSHARED_HEK_FROM_PV(SvPVX_const(sstr)))));
bdd68bc3 4141 }
87a1ef3d
SP
4142 SvLEN_set(dstr, len);
4143 SvCUR_set(dstr, cur);
765f542d
NC
4144 SvREADONLY_on(dstr);
4145 SvFAKE_on(dstr);
765f542d
NC
4146 }
4147 else
765f542d 4148 { /* Passes the swipe test. */
78d1e721 4149 SvPV_set(dstr, SvPVX_mutable(sstr));
765f542d
NC
4150 SvLEN_set(dstr, SvLEN(sstr));
4151 SvCUR_set(dstr, SvCUR(sstr));
4152
4153 SvTEMP_off(dstr);
4154 (void)SvOK_off(sstr); /* NOTE: nukes most SvFLAGS on sstr */
6136c704 4155 SvPV_set(sstr, NULL);
765f542d
NC
4156 SvLEN_set(sstr, 0);
4157 SvCUR_set(sstr, 0);
4158 SvTEMP_off(sstr);
4159 }
4160 }
8990e307 4161 if (sflags & SVp_NOK) {
9d6ce603 4162 SvNV_set(dstr, SvNVX(sstr));
79072805 4163 }
8990e307 4164 if (sflags & SVp_IOK) {
23525414
NC
4165 SvIV_set(dstr, SvIVX(sstr));
4166 /* Must do this otherwise some other overloaded use of 0x80000000
4167 gets confused. I guess SVpbm_VALID */
2b1c7e3e 4168 if (sflags & SVf_IVisUV)
25da4f38 4169 SvIsUV_on(dstr);
79072805 4170 }
96d4b0ee 4171 SvFLAGS(dstr) |= sflags & (SVf_IOK|SVp_IOK|SVf_NOK|SVp_NOK|SVf_UTF8);
4f2da183 4172 {
b0a11fe1 4173 const MAGIC * const smg = SvVSTRING_mg(sstr);
4f2da183
NC
4174 if (smg) {
4175 sv_magic(dstr, NULL, PERL_MAGIC_vstring,
4176 smg->mg_ptr, smg->mg_len);
4177 SvRMAGICAL_on(dstr);
4178 }
7a5fa8a2 4179 }
79072805 4180 }
5d581361 4181 else if (sflags & (SVp_IOK|SVp_NOK)) {
c2468cc7 4182 (void)SvOK_off(dstr);
96d4b0ee 4183 SvFLAGS(dstr) |= sflags & (SVf_IOK|SVp_IOK|SVf_IVisUV|SVf_NOK|SVp_NOK);
5d581361
NC
4184 if (sflags & SVp_IOK) {
4185 /* XXXX Do we want to set IsUV for IV(ROK)? Be extra safe... */
4186 SvIV_set(dstr, SvIVX(sstr));
4187 }
3332b3c1 4188 if (sflags & SVp_NOK) {
9d6ce603 4189 SvNV_set(dstr, SvNVX(sstr));
3332b3c1
JH
4190 }
4191 }
79072805 4192 else {
f7877b28 4193 if (isGV_with_GP(sstr)) {
180488f8
NC
4194 /* This stringification rule for globs is spread in 3 places.
4195 This feels bad. FIXME. */
4196 const U32 wasfake = sflags & SVf_FAKE;
4197
4198 /* FAKE globs can get coerced, so need to turn this off
4199 temporarily if it is on. */
4200 SvFAKE_off(sstr);
159b6efe 4201 gv_efullname3(dstr, MUTABLE_GV(sstr), "*");
180488f8
NC
4202 SvFLAGS(sstr) |= wasfake;
4203 }
20408e3c
GS
4204 else
4205 (void)SvOK_off(dstr);
a0d0e21e 4206 }
27c9684d
AP
4207 if (SvTAINTED(sstr))
4208 SvTAINT(dstr);
79072805
LW
4209}
4210
954c1994
GS
4211/*
4212=for apidoc sv_setsv_mg
4213
4214Like C<sv_setsv>, but also handles 'set' magic.
4215
4216=cut
4217*/
4218
79072805 4219void
7bc54cea 4220Perl_sv_setsv_mg(pTHX_ SV *const dstr, register SV *const sstr)
ef50df4b 4221{
7918f24d
NC
4222 PERL_ARGS_ASSERT_SV_SETSV_MG;
4223
ef50df4b
GS
4224 sv_setsv(dstr,sstr);
4225 SvSETMAGIC(dstr);
4226}
4227
f8c7b90f 4228#ifdef PERL_OLD_COPY_ON_WRITE
ed252734
NC
4229SV *
4230Perl_sv_setsv_cow(pTHX_ SV *dstr, SV *sstr)
4231{
4232 STRLEN cur = SvCUR(sstr);
4233 STRLEN len = SvLEN(sstr);
4234 register char *new_pv;
4235
7918f24d
NC
4236 PERL_ARGS_ASSERT_SV_SETSV_COW;
4237
ed252734
NC
4238 if (DEBUG_C_TEST) {
4239 PerlIO_printf(Perl_debug_log, "Fast copy on write: %p -> %p\n",
6c9570dc 4240 (void*)sstr, (void*)dstr);
ed252734
NC
4241 sv_dump(sstr);
4242 if (dstr)
4243 sv_dump(dstr);
4244 }
4245
4246 if (dstr) {
4247 if (SvTHINKFIRST(dstr))
4248 sv_force_normal_flags(dstr, SV_COW_DROP_PV);
3f7c398e
SP
4249 else if (SvPVX_const(dstr))
4250 Safefree(SvPVX_const(dstr));
ed252734
NC
4251 }
4252 else
4253 new_SV(dstr);
862a34c6 4254 SvUPGRADE(dstr, SVt_PVIV);
ed252734
NC
4255
4256 assert (SvPOK(sstr));
4257 assert (SvPOKp(sstr));
4258 assert (!SvIOK(sstr));
4259 assert (!SvIOKp(sstr));
4260 assert (!SvNOK(sstr));
4261 assert (!SvNOKp(sstr));
4262
4263 if (SvIsCOW(sstr)) {
4264
4265 if (SvLEN(sstr) == 0) {
4266 /* source is a COW shared hash key. */
ed252734
NC
4267 DEBUG_C(PerlIO_printf(Perl_debug_log,
4268 "Fast copy on write: Sharing hash\n"));
d1db91c6 4269 new_pv = HEK_KEY(share_hek_hek(SvSHARED_HEK_FROM_PV(SvPVX_const(sstr))));
ed252734
NC
4270 goto common_exit;
4271 }
4272 SV_COW_NEXT_SV_SET(dstr, SV_COW_NEXT_SV(sstr));
4273 } else {
4274 assert ((SvFLAGS(sstr) & CAN_COW_MASK) == CAN_COW_FLAGS);
862a34c6 4275 SvUPGRADE(sstr, SVt_PVIV);
ed252734
NC
4276 SvREADONLY_on(sstr);
4277 SvFAKE_on(sstr);
4278 DEBUG_C(PerlIO_printf(Perl_debug_log,
4279 "Fast copy on write: Converting sstr to COW\n"));
4280 SV_COW_NEXT_SV_SET(dstr, sstr);
4281 }
4282 SV_COW_NEXT_SV_SET(sstr, dstr);
940132f3 4283 new_pv = SvPVX_mutable(sstr);
ed252734
NC
4284
4285 common_exit:
4286 SvPV_set(dstr, new_pv);
4287 SvFLAGS(dstr) = (SVt_PVIV|SVf_POK|SVp_POK|SVf_FAKE|SVf_READONLY);
4288 if (SvUTF8(sstr))
4289 SvUTF8_on(dstr);
87a1ef3d
SP
4290 SvLEN_set(dstr, len);
4291 SvCUR_set(dstr, cur);
ed252734
NC
4292 if (DEBUG_C_TEST) {
4293 sv_dump(dstr);
4294 }
4295 return dstr;
4296}
4297#endif
4298
954c1994
GS
4299/*
4300=for apidoc sv_setpvn
4301
4302Copies a string into an SV. The C<len> parameter indicates the number of
9e09f5f2
MHM
4303bytes to be copied. If the C<ptr> argument is NULL the SV will become
4304undefined. Does not handle 'set' magic. See C<sv_setpvn_mg>.
954c1994
GS
4305
4306=cut
4307*/
4308
ef50df4b 4309void
2e000ff2 4310Perl_sv_setpvn(pTHX_ register SV *const sv, register const char *const ptr, register const STRLEN len)
79072805 4311{
97aff369 4312 dVAR;
c6f8c383 4313 register char *dptr;
22c522df 4314
7918f24d
NC
4315 PERL_ARGS_ASSERT_SV_SETPVN;
4316
765f542d 4317 SV_CHECK_THINKFIRST_COW_DROP(sv);
463ee0b2 4318 if (!ptr) {
a0d0e21e 4319 (void)SvOK_off(sv);
463ee0b2
LW
4320 return;
4321 }
22c522df
JH
4322 else {
4323 /* len is STRLEN which is unsigned, need to copy to signed */
a3b680e6 4324 const IV iv = len;
9c5ffd7c
JH
4325 if (iv < 0)
4326 Perl_croak(aTHX_ "panic: sv_setpvn called with negative strlen");
22c522df 4327 }
862a34c6 4328 SvUPGRADE(sv, SVt_PV);
c6f8c383 4329
5902b6a9 4330 dptr = SvGROW(sv, len + 1);
c6f8c383
GA
4331 Move(ptr,dptr,len,char);
4332 dptr[len] = '\0';
79072805 4333 SvCUR_set(sv, len);
1aa99e6b 4334 (void)SvPOK_only_UTF8(sv); /* validate pointer */
463ee0b2 4335 SvTAINT(sv);
79072805
LW
4336}
4337
954c1994
GS
4338/*
4339=for apidoc sv_setpvn_mg
4340
4341Like C<sv_setpvn>, but also handles 'set' magic.
4342
4343=cut
4344*/
4345
79072805 4346void
2e000ff2 4347Perl_sv_setpvn_mg(pTHX_ register SV *const sv, register const char *const ptr, register const STRLEN len)
ef50df4b 4348{
7918f24d
NC
4349 PERL_ARGS_ASSERT_SV_SETPVN_MG;
4350
ef50df4b
GS
4351 sv_setpvn(sv,ptr,len);
4352 SvSETMAGIC(sv);
4353}
4354
954c1994
GS
4355/*
4356=for apidoc sv_setpv
4357
4358Copies a string into an SV. The string must be null-terminated. Does not
4359handle 'set' magic. See C<sv_setpv_mg>.
4360
4361=cut
4362*/
4363
ef50df4b 4364void
2e000ff2 4365Perl_sv_setpv(pTHX_ register SV *const sv, register const char *const ptr)
79072805 4366{
97aff369 4367 dVAR;
79072805
LW
4368 register STRLEN len;
4369
7918f24d
NC
4370 PERL_ARGS_ASSERT_SV_SETPV;
4371
765f542d 4372 SV_CHECK_THINKFIRST_COW_DROP(sv);
463ee0b2 4373 if (!ptr) {
a0d0e21e 4374 (void)SvOK_off(sv);
463ee0b2
LW
4375 return;
4376 }
79072805 4377 len = strlen(ptr);
862a34c6 4378 SvUPGRADE(sv, SVt_PV);
c6f8c383 4379
79072805 4380 SvGROW(sv, len + 1);
463ee0b2 4381 Move(ptr,SvPVX(sv),len+1,char);
79072805 4382 SvCUR_set(sv, len);
1aa99e6b 4383 (void)SvPOK_only_UTF8(sv); /* validate pointer */
463ee0b2
LW
4384 SvTAINT(sv);
4385}
4386
954c1994
GS
4387/*
4388=for apidoc sv_setpv_mg
4389
4390Like C<sv_setpv>, but also handles 'set' magic.
4391
4392=cut
4393*/
4394
463ee0b2 4395void
2e000ff2 4396Perl_sv_setpv_mg(pTHX_ register SV *const sv, register const char *const ptr)
ef50df4b 4397{
7918f24d
NC
4398 PERL_ARGS_ASSERT_SV_SETPV_MG;
4399
ef50df4b
GS
4400 sv_setpv(sv,ptr);
4401 SvSETMAGIC(sv);
4402}
4403
954c1994 4404/*
47518d95 4405=for apidoc sv_usepvn_flags
954c1994 4406
794a0d33
JH
4407Tells an SV to use C<ptr> to find its string value. Normally the
4408string is stored inside the SV but sv_usepvn allows the SV to use an
4409outside string. The C<ptr> should point to memory that was allocated
c1c21316
NC
4410by C<malloc>. The string length, C<len>, must be supplied. By default
4411this function will realloc (i.e. move) the memory pointed to by C<ptr>,
794a0d33
JH
4412so that pointer should not be freed or used by the programmer after
4413giving it to sv_usepvn, and neither should any pointers from "behind"
c1c21316
NC
4414that pointer (e.g. ptr + 1) be used.
4415
4416If C<flags> & SV_SMAGIC is true, will call SvSETMAGIC. If C<flags> &
4417SV_HAS_TRAILING_NUL is true, then C<ptr[len]> must be NUL, and the realloc
cbf82dd0 4418will be skipped. (i.e. the buffer is actually at least 1 byte longer than
c1c21316 4419C<len>, and already meets the requirements for storing in C<SvPVX>)
954c1994
GS
4420
4421=cut
4422*/
4423
ef50df4b 4424void
2e000ff2 4425Perl_sv_usepvn_flags(pTHX_ SV *const sv, char *ptr, const STRLEN len, const U32 flags)
463ee0b2 4426{
97aff369 4427 dVAR;
1936d2a7 4428 STRLEN allocate;
7918f24d
NC
4429
4430 PERL_ARGS_ASSERT_SV_USEPVN_FLAGS;
4431
765f542d 4432 SV_CHECK_THINKFIRST_COW_DROP(sv);
862a34c6 4433 SvUPGRADE(sv, SVt_PV);
463ee0b2 4434 if (!ptr) {
a0d0e21e 4435 (void)SvOK_off(sv);
47518d95
NC
4436 if (flags & SV_SMAGIC)
4437 SvSETMAGIC(sv);
463ee0b2
LW
4438 return;
4439 }
3f7c398e 4440 if (SvPVX_const(sv))
8bd4d4c5 4441 SvPV_free(sv);
1936d2a7 4442
0b7042f9 4443#ifdef DEBUGGING
2e90b4cd
NC
4444 if (flags & SV_HAS_TRAILING_NUL)
4445 assert(ptr[len] == '\0');
0b7042f9 4446#endif
2e90b4cd 4447
c1c21316 4448 allocate = (flags & SV_HAS_TRAILING_NUL)
5d487c26 4449 ? len + 1 :
ca7c1a29 4450#ifdef Perl_safesysmalloc_size
5d487c26
NC
4451 len + 1;
4452#else
4453 PERL_STRLEN_ROUNDUP(len + 1);
4454#endif
cbf82dd0
NC
4455 if (flags & SV_HAS_TRAILING_NUL) {
4456 /* It's long enough - do nothing.
4457 Specfically Perl_newCONSTSUB is relying on this. */
4458 } else {
69d25b4f 4459#ifdef DEBUGGING
69d25b4f 4460 /* Force a move to shake out bugs in callers. */
10edeb5d 4461 char *new_ptr = (char*)safemalloc(allocate);
69d25b4f
NC
4462 Copy(ptr, new_ptr, len, char);
4463 PoisonFree(ptr,len,char);
4464 Safefree(ptr);
4465 ptr = new_ptr;
69d25b4f 4466#else
10edeb5d 4467 ptr = (char*) saferealloc (ptr, allocate);
69d25b4f 4468#endif
cbf82dd0 4469 }
ca7c1a29
NC
4470#ifdef Perl_safesysmalloc_size
4471 SvLEN_set(sv, Perl_safesysmalloc_size(ptr));
5d487c26 4472#else
1936d2a7 4473 SvLEN_set(sv, allocate);
5d487c26
NC
4474#endif
4475 SvCUR_set(sv, len);
4476 SvPV_set(sv, ptr);
c1c21316 4477 if (!(flags & SV_HAS_TRAILING_NUL)) {
97a130b8 4478 ptr[len] = '\0';
c1c21316 4479 }
1aa99e6b 4480 (void)SvPOK_only_UTF8(sv); /* validate pointer */
463ee0b2 4481 SvTAINT(sv);
47518d95
NC
4482 if (flags & SV_SMAGIC)
4483 SvSETMAGIC(sv);
ef50df4b
GS
4484}
4485
f8c7b90f 4486#ifdef PERL_OLD_COPY_ON_WRITE
765f542d
NC
4487/* Need to do this *after* making the SV normal, as we need the buffer
4488 pointer to remain valid until after we've copied it. If we let go too early,
4489 another thread could invalidate it by unsharing last of the same hash key
4490 (which it can do by means other than releasing copy-on-write Svs)
4491 or by changing the other copy-on-write SVs in the loop. */
4492STATIC void
5302ffd4 4493S_sv_release_COW(pTHX_ register SV *sv, const char *pvx, SV *after)
765f542d 4494{
7918f24d
NC
4495 PERL_ARGS_ASSERT_SV_RELEASE_COW;
4496
5302ffd4 4497 { /* this SV was SvIsCOW_normal(sv) */
765f542d 4498 /* we need to find the SV pointing to us. */
cf5629ad 4499 SV *current = SV_COW_NEXT_SV(after);
7a5fa8a2 4500
765f542d
NC
4501 if (current == sv) {
4502 /* The SV we point to points back to us (there were only two of us
4503 in the loop.)
4504 Hence other SV is no longer copy on write either. */
4505 SvFAKE_off(after);
4506 SvREADONLY_off(after);
4507 } else {
4508 /* We need to follow the pointers around the loop. */
4509 SV *next;
4510 while ((next = SV_COW_NEXT_SV(current)) != sv) {
4511 assert (next);
4512 current = next;
4513 /* don't loop forever if the structure is bust, and we have
4514 a pointer into a closed loop. */
4515 assert (current != after);
3f7c398e 4516 assert (SvPVX_const(current) == pvx);
765f542d
NC
4517 }
4518 /* Make the SV before us point to the SV after us. */
a29f6d03 4519 SV_COW_NEXT_SV_SET(current, after);
765f542d 4520 }
765f542d
NC
4521 }
4522}
765f542d 4523#endif
645c22ef
DM
4524/*
4525=for apidoc sv_force_normal_flags
4526
4527Undo various types of fakery on an SV: if the PV is a shared string, make
4528a private copy; if we're a ref, stop refing; if we're a glob, downgrade to
765f542d
NC
4529an xpvmg; if we're a copy-on-write scalar, this is the on-write time when
4530we do the copy, and is also used locally. If C<SV_COW_DROP_PV> is set
4531then a copy-on-write scalar drops its PV buffer (if any) and becomes
4532SvPOK_off rather than making a copy. (Used where this scalar is about to be
d3050d9d 4533set to some other value.) In addition, the C<flags> parameter gets passed to
765f542d
NC
4534C<sv_unref_flags()> when unrefing. C<sv_force_normal> calls this function
4535with flags set to 0.
645c22ef
DM
4536
4537=cut
4538*/
4539
6fc92669 4540void
2e000ff2 4541Perl_sv_force_normal_flags(pTHX_ register SV *const sv, const U32 flags)
0f15f207 4542{
97aff369 4543 dVAR;
7918f24d
NC
4544
4545 PERL_ARGS_ASSERT_SV_FORCE_NORMAL_FLAGS;
4546
f8c7b90f 4547#ifdef PERL_OLD_COPY_ON_WRITE
765f542d 4548 if (SvREADONLY(sv)) {
765f542d 4549 if (SvFAKE(sv)) {
b64e5050 4550 const char * const pvx = SvPVX_const(sv);
a28509cc
AL
4551 const STRLEN len = SvLEN(sv);
4552 const STRLEN cur = SvCUR(sv);
5302ffd4
NC
4553 /* next COW sv in the loop. If len is 0 then this is a shared-hash
4554 key scalar, so we mustn't attempt to call SV_COW_NEXT_SV(), as
4555 we'll fail an assertion. */
4556 SV * const next = len ? SV_COW_NEXT_SV(sv) : 0;
4557
46187eeb
NC
4558 if (DEBUG_C_TEST) {
4559 PerlIO_printf(Perl_debug_log,
4560 "Copy on write: Force normal %ld\n",
4561 (long) flags);
e419cbc5 4562 sv_dump(sv);
46187eeb 4563 }
765f542d
NC
4564 SvFAKE_off(sv);
4565 SvREADONLY_off(sv);
9f653bb5 4566 /* This SV doesn't own the buffer, so need to Newx() a new one: */
6136c704 4567 SvPV_set(sv, NULL);
87a1ef3d 4568 SvLEN_set(sv, 0);
765f542d
NC
4569 if (flags & SV_COW_DROP_PV) {
4570 /* OK, so we don't need to copy our buffer. */
4571 SvPOK_off(sv);
4572 } else {
4573 SvGROW(sv, cur + 1);
4574 Move(pvx,SvPVX(sv),cur,char);
87a1ef3d 4575 SvCUR_set(sv, cur);
765f542d
NC
4576 *SvEND(sv) = '\0';
4577 }
5302ffd4
NC
4578 if (len) {
4579 sv_release_COW(sv, pvx, next);
4580 } else {
4581 unshare_hek(SvSHARED_HEK_FROM_PV(pvx));
4582 }
46187eeb 4583 if (DEBUG_C_TEST) {
e419cbc5 4584 sv_dump(sv);
46187eeb 4585 }
765f542d 4586 }
923e4eb5 4587 else if (IN_PERL_RUNTIME)
f1f66076 4588 Perl_croak(aTHX_ "%s", PL_no_modify);
765f542d
NC
4589 }
4590#else
2213622d 4591 if (SvREADONLY(sv)) {
1c846c1f 4592 if (SvFAKE(sv)) {
b64e5050 4593 const char * const pvx = SvPVX_const(sv);
66a1b24b 4594 const STRLEN len = SvCUR(sv);
10bcdfd6
NC
4595 SvFAKE_off(sv);
4596 SvREADONLY_off(sv);
bd61b366 4597 SvPV_set(sv, NULL);
66a1b24b 4598 SvLEN_set(sv, 0);
1c846c1f 4599 SvGROW(sv, len + 1);
706aa1c9 4600 Move(pvx,SvPVX(sv),len,char);
1c846c1f 4601 *SvEND(sv) = '\0';
bdd68bc3 4602 unshare_hek(SvSHARED_HEK_FROM_PV(pvx));
1c846c1f 4603 }
923e4eb5 4604 else if (IN_PERL_RUNTIME)
f1f66076 4605 Perl_croak(aTHX_ "%s", PL_no_modify);
0f15f207 4606 }
765f542d 4607#endif
2213622d 4608 if (SvROK(sv))
840a7b70 4609 sv_unref_flags(sv, flags);
6fc92669
GS
4610 else if (SvFAKE(sv) && SvTYPE(sv) == SVt_PVGV)
4611 sv_unglob(sv);
0f15f207 4612}
1c846c1f 4613
645c22ef 4614/*
954c1994
GS
4615=for apidoc sv_chop
4616
1c846c1f 4617Efficient removal of characters from the beginning of the string buffer.
954c1994
GS
4618SvPOK(sv) must be true and the C<ptr> must be a pointer to somewhere inside
4619the string buffer. The C<ptr> becomes the first character of the adjusted
645c22ef 4620string. Uses the "OOK hack".
3f7c398e 4621Beware: after this function returns, C<ptr> and SvPVX_const(sv) may no longer
31869a79 4622refer to the same chunk of data.
954c1994
GS
4623
4624=cut
4625*/
4626
79072805 4627void
2e000ff2 4628Perl_sv_chop(pTHX_ register SV *const sv, register const char *const ptr)
79072805 4629{
69240efd
NC
4630 STRLEN delta;
4631 STRLEN old_delta;
7a4bba22
NC
4632 U8 *p;
4633#ifdef DEBUGGING
4634 const U8 *real_start;
4635#endif
6c65d5f9 4636 STRLEN max_delta;
7a4bba22 4637
7918f24d
NC
4638 PERL_ARGS_ASSERT_SV_CHOP;
4639
a0d0e21e 4640 if (!ptr || !SvPOKp(sv))
79072805 4641 return;
3f7c398e 4642 delta = ptr - SvPVX_const(sv);
15895f8a
NC
4643 if (!delta) {
4644 /* Nothing to do. */
4645 return;
4646 }
6c65d5f9
NC
4647 /* SvPVX(sv) may move in SV_CHECK_THINKFIRST(sv), but after this line,
4648 nothing uses the value of ptr any more. */
837cb3ba 4649 max_delta = SvLEN(sv) ? SvLEN(sv) : SvCUR(sv);
6c65d5f9
NC
4650 if (ptr <= SvPVX_const(sv))
4651 Perl_croak(aTHX_ "panic: sv_chop ptr=%p, start=%p, end=%p",
4652 ptr, SvPVX_const(sv), SvPVX_const(sv) + max_delta);
2213622d 4653 SV_CHECK_THINKFIRST(sv);
6c65d5f9
NC
4654 if (delta > max_delta)
4655 Perl_croak(aTHX_ "panic: sv_chop ptr=%p (was %p), start=%p, end=%p",
4656 SvPVX_const(sv) + delta, ptr, SvPVX_const(sv),
4657 SvPVX_const(sv) + max_delta);
79072805
LW
4658
4659 if (!SvOOK(sv)) {
50483b2c 4660 if (!SvLEN(sv)) { /* make copy of shared string */
3f7c398e 4661 const char *pvx = SvPVX_const(sv);
a28509cc 4662 const STRLEN len = SvCUR(sv);
50483b2c 4663 SvGROW(sv, len + 1);
706aa1c9 4664 Move(pvx,SvPVX(sv),len,char);
50483b2c
JD
4665 *SvEND(sv) = '\0';
4666 }
7a5fa8a2 4667 SvFLAGS(sv) |= SVf_OOK;
7a4bba22
NC
4668 old_delta = 0;
4669 } else {
69240efd 4670 SvOOK_offset(sv, old_delta);
79072805 4671 }
b162af07
SP
4672 SvLEN_set(sv, SvLEN(sv) - delta);
4673 SvCUR_set(sv, SvCUR(sv) - delta);
f880fe2f 4674 SvPV_set(sv, SvPVX(sv) + delta);
7a4bba22
NC
4675
4676 p = (U8 *)SvPVX_const(sv);
4677
4678 delta += old_delta;
4679
50af2e61 4680#ifdef DEBUGGING
7a4bba22
NC
4681 real_start = p - delta;
4682#endif
4683
69240efd
NC
4684 assert(delta);
4685 if (delta < 0x100) {
7a4bba22
NC
4686 *--p = (U8) delta;
4687 } else {
69240efd
NC
4688 *--p = 0;
4689 p -= sizeof(STRLEN);
4690 Copy((U8*)&delta, p, sizeof(STRLEN), U8);
7a4bba22
NC
4691 }
4692
4693#ifdef DEBUGGING
4694 /* Fill the preceding buffer with sentinals to verify that no-one is
4695 using it. */
4696 while (p > real_start) {
4697 --p;
4698 *p = (U8)PTR2UV(p);
50af2e61
NC
4699 }
4700#endif
79072805
LW
4701}
4702
954c1994
GS
4703/*
4704=for apidoc sv_catpvn
4705
4706Concatenates the string onto the end of the string which is in the SV. The
1e54db1a
JH
4707C<len> indicates number of bytes to copy. If the SV has the UTF-8
4708status set, then the bytes appended should be valid UTF-8.
d5ce4a7c 4709Handles 'get' magic, but not 'set' magic. See C<sv_catpvn_mg>.
954c1994 4710
8d6d96c1
HS
4711=for apidoc sv_catpvn_flags
4712
4713Concatenates the string onto the end of the string which is in the SV. The
1e54db1a
JH
4714C<len> indicates number of bytes to copy. If the SV has the UTF-8
4715status set, then the bytes appended should be valid UTF-8.
8d6d96c1
HS
4716If C<flags> has C<SV_GMAGIC> bit set, will C<mg_get> on C<dsv> if
4717appropriate, else not. C<sv_catpvn> and C<sv_catpvn_nomg> are implemented
4718in terms of this function.
4719
4720=cut
4721*/
4722
4723void
2e000ff2 4724Perl_sv_catpvn_flags(pTHX_ register SV *const dsv, register const char *sstr, register const STRLEN slen, const I32 flags)
8d6d96c1 4725{
97aff369 4726 dVAR;
8d6d96c1 4727 STRLEN dlen;
fabdb6c0 4728 const char * const dstr = SvPV_force_flags(dsv, dlen, flags);
8d6d96c1 4729
7918f24d
NC
4730 PERL_ARGS_ASSERT_SV_CATPVN_FLAGS;
4731
8d6d96c1
HS
4732 SvGROW(dsv, dlen + slen + 1);
4733 if (sstr == dstr)
3f7c398e 4734 sstr = SvPVX_const(dsv);
8d6d96c1 4735 Move(sstr, SvPVX(dsv) + dlen, slen, char);
b162af07 4736 SvCUR_set(dsv, SvCUR(dsv) + slen);
8d6d96c1
HS
4737 *SvEND(dsv) = '\0';
4738 (void)SvPOK_only_UTF8(dsv); /* validate pointer */
4739 SvTAINT(dsv);
bddd5118
NC
4740 if (flags & SV_SMAGIC)
4741 SvSETMAGIC(dsv);
79072805
LW
4742}
4743
954c1994 4744/*
954c1994
GS
4745=for apidoc sv_catsv
4746
13e8c8e3
JH
4747Concatenates the string from SV C<ssv> onto the end of the string in
4748SV C<dsv>. Modifies C<dsv> but not C<ssv>. Handles 'get' magic, but
4749not 'set' magic. See C<sv_catsv_mg>.
954c1994 4750
8d6d96c1
HS
4751=for apidoc sv_catsv_flags
4752
4753Concatenates the string from SV C<ssv> onto the end of the string in
4754SV C<dsv>. Modifies C<dsv> but not C<ssv>. If C<flags> has C<SV_GMAGIC>
4755bit set, will C<mg_get> on the SVs if appropriate, else not. C<sv_catsv>
4756and C<sv_catsv_nomg> are implemented in terms of this function.
4757
4758=cut */
4759
ef50df4b 4760void
2e000ff2 4761Perl_sv_catsv_flags(pTHX_ SV *const dsv, register SV *const ssv, const I32 flags)
79072805 4762{
97aff369 4763 dVAR;
7918f24d
NC
4764
4765 PERL_ARGS_ASSERT_SV_CATSV_FLAGS;
4766
4767 if (ssv) {
00b6aa41
AL
4768 STRLEN slen;
4769 const char *spv = SvPV_const(ssv, slen);
4770 if (spv) {
bddd5118
NC
4771 /* sutf8 and dutf8 were type bool, but under USE_ITHREADS,
4772 gcc version 2.95.2 20000220 (Debian GNU/Linux) for
4773 Linux xxx 2.2.17 on sparc64 with gcc -O2, we erroneously
4774 get dutf8 = 0x20000000, (i.e. SVf_UTF8) even though
4775 dsv->sv_flags doesn't have that bit set.
4fd84b44 4776 Andy Dougherty 12 Oct 2001
bddd5118
NC
4777 */
4778 const I32 sutf8 = DO_UTF8(ssv);
4779 I32 dutf8;
13e8c8e3 4780
bddd5118
NC
4781 if (SvGMAGICAL(dsv) && (flags & SV_GMAGIC))
4782 mg_get(dsv);
4783 dutf8 = DO_UTF8(dsv);
8d6d96c1 4784
bddd5118
NC
4785 if (dutf8 != sutf8) {
4786 if (dutf8) {
4787 /* Not modifying source SV, so taking a temporary copy. */
59cd0e26 4788 SV* const csv = newSVpvn_flags(spv, slen, SVs_TEMP);
13e8c8e3 4789
bddd5118
NC
4790 sv_utf8_upgrade(csv);
4791 spv = SvPV_const(csv, slen);
4792 }
4793 else
7bf79863
KW
4794 /* Leave enough space for the cat that's about to happen */
4795 sv_utf8_upgrade_flags_grow(dsv, 0, slen);
13e8c8e3 4796 }
bddd5118 4797 sv_catpvn_nomg(dsv, spv, slen);
e84ff256 4798 }
560a288e 4799 }
bddd5118
NC
4800 if (flags & SV_SMAGIC)
4801 SvSETMAGIC(dsv);
79072805
LW
4802}
4803
954c1994 4804/*
954c1994
GS
4805=for apidoc sv_catpv
4806
4807Concatenates the string onto the end of the string which is in the SV.
1e54db1a
JH
4808If the SV has the UTF-8 status set, then the bytes appended should be
4809valid UTF-8. Handles 'get' magic, but not 'set' magic. See C<sv_catpv_mg>.
954c1994 4810
d5ce4a7c 4811=cut */
954c1994 4812
ef50df4b 4813void
2b021c53 4814Perl_sv_catpv(pTHX_ register SV *const sv, register const char *ptr)
79072805 4815{
97aff369 4816 dVAR;
79072805 4817 register STRLEN len;
463ee0b2 4818 STRLEN tlen;
748a9306 4819 char *junk;
79072805 4820
7918f24d
NC
4821 PERL_ARGS_ASSERT_SV_CATPV;
4822
0c981600 4823 if (!ptr)
79072805 4824 return;
748a9306 4825 junk = SvPV_force(sv, tlen);
0c981600 4826 len = strlen(ptr);
463ee0b2 4827 SvGROW(sv, tlen + len + 1);
0c981600 4828 if (ptr == junk)
3f7c398e 4829 ptr = SvPVX_const(sv);
0c981600 4830 Move(ptr,SvPVX(sv)+tlen,len+1,char);
b162af07 4831 SvCUR_set(sv, SvCUR(sv) + len);
d41ff1b8 4832 (void)SvPOK_only_UTF8(sv); /* validate pointer */
463ee0b2 4833 SvTAINT(sv);
79072805
LW
4834}
4835
954c1994
GS
4836/*
4837=for apidoc sv_catpv_mg
4838
4839Like C<sv_catpv>, but also handles 'set' magic.
4840
4841=cut
4842*/
4843
ef50df4b 4844void
2b021c53 4845Perl_sv_catpv_mg(pTHX_ register SV *const sv, register const char *const ptr)
ef50df4b 4846{
7918f24d
NC
4847 PERL_ARGS_ASSERT_SV_CATPV_MG;
4848
0c981600 4849 sv_catpv(sv,ptr);
ef50df4b
GS
4850 SvSETMAGIC(sv);
4851}
4852
645c22ef
DM
4853/*
4854=for apidoc newSV
4855
561b68a9
SH
4856Creates a new SV. A non-zero C<len> parameter indicates the number of
4857bytes of preallocated string space the SV should have. An extra byte for a
4858trailing NUL is also reserved. (SvPOK is not set for the SV even if string
4859space is allocated.) The reference count for the new SV is set to 1.
4860
4861In 5.9.3, newSV() replaces the older NEWSV() API, and drops the first
4862parameter, I<x>, a debug aid which allowed callers to identify themselves.
4863This aid has been superseded by a new build option, PERL_MEM_LOG (see
4864L<perlhack/PERL_MEM_LOG>). The older API is still there for use in XS
4865modules supporting older perls.
645c22ef
DM
4866
4867=cut
4868*/
4869
79072805 4870SV *
2b021c53 4871Perl_newSV(pTHX_ const STRLEN len)
79072805 4872{
97aff369 4873 dVAR;
79072805 4874 register SV *sv;
1c846c1f 4875
4561caa4 4876 new_SV(sv);
79072805
LW
4877 if (len) {
4878 sv_upgrade(sv, SVt_PV);
4879 SvGROW(sv, len + 1);
4880 }
4881 return sv;
4882}
954c1994 4883/*
92110913 4884=for apidoc sv_magicext
954c1994 4885
68795e93 4886Adds magic to an SV, upgrading it if necessary. Applies the
2d8d5d5a 4887supplied vtable and returns a pointer to the magic added.
92110913 4888
2d8d5d5a
SH
4889Note that C<sv_magicext> will allow things that C<sv_magic> will not.
4890In particular, you can add magic to SvREADONLY SVs, and add more than
4891one instance of the same 'how'.
645c22ef 4892
2d8d5d5a
SH
4893If C<namlen> is greater than zero then a C<savepvn> I<copy> of C<name> is
4894stored, if C<namlen> is zero then C<name> is stored as-is and - as another
4895special case - if C<(name && namlen == HEf_SVKEY)> then C<name> is assumed
4896to contain an C<SV*> and is stored as-is with its REFCNT incremented.
92110913 4897
2d8d5d5a 4898(This is now used as a subroutine by C<sv_magic>.)
954c1994
GS
4899
4900=cut
4901*/
92110913 4902MAGIC *
2b021c53
SS
4903Perl_sv_magicext(pTHX_ SV *const sv, SV *const obj, const int how,
4904 const MGVTBL *const vtable, const char *const name, const I32 namlen)
79072805 4905{
97aff369 4906 dVAR;
79072805 4907 MAGIC* mg;
68795e93 4908
7918f24d
NC
4909 PERL_ARGS_ASSERT_SV_MAGICEXT;
4910
7a7f3e45 4911 SvUPGRADE(sv, SVt_PVMG);
a02a5408 4912 Newxz(mg, 1, MAGIC);
79072805 4913 mg->mg_moremagic = SvMAGIC(sv);
b162af07 4914 SvMAGIC_set(sv, mg);
75f9d97a 4915
05f95b08
SB
4916 /* Sometimes a magic contains a reference loop, where the sv and
4917 object refer to each other. To prevent a reference loop that
4918 would prevent such objects being freed, we look for such loops
4919 and if we find one we avoid incrementing the object refcount.
87f0b213
JH
4920
4921 Note we cannot do this to avoid self-tie loops as intervening RV must
b5ccf5f2 4922 have its REFCNT incremented to keep it in existence.
87f0b213
JH
4923
4924 */
14befaf4
DM
4925 if (!obj || obj == sv ||
4926 how == PERL_MAGIC_arylen ||
8d2f4536 4927 how == PERL_MAGIC_symtab ||
75f9d97a 4928 (SvTYPE(obj) == SVt_PVGV &&
4c4652b6
NC
4929 (GvSV(obj) == sv || GvHV(obj) == (const HV *)sv
4930 || GvAV(obj) == (const AV *)sv || GvCV(obj) == (const CV *)sv
4931 || GvIOp(obj) == (const IO *)sv || GvFORM(obj) == (const CV *)sv)))
75f9d97a 4932 {
8990e307 4933 mg->mg_obj = obj;
75f9d97a 4934 }
85e6fe83 4935 else {
b37c2d43 4936 mg->mg_obj = SvREFCNT_inc_simple(obj);
85e6fe83
LW
4937 mg->mg_flags |= MGf_REFCOUNTED;
4938 }
b5ccf5f2
YST
4939
4940 /* Normal self-ties simply pass a null object, and instead of
4941 using mg_obj directly, use the SvTIED_obj macro to produce a
4942 new RV as needed. For glob "self-ties", we are tieing the PVIO
4943 with an RV obj pointing to the glob containing the PVIO. In
4944 this case, to avoid a reference loop, we need to weaken the
4945 reference.
4946 */
4947
4948 if (how == PERL_MAGIC_tiedscalar && SvTYPE(sv) == SVt_PVIO &&
a45c7426 4949 obj && SvROK(obj) && GvIO(SvRV(obj)) == (const IO *)sv)
b5ccf5f2
YST
4950 {
4951 sv_rvweaken(obj);
4952 }
4953
79072805 4954 mg->mg_type = how;
565764a8 4955 mg->mg_len = namlen;
9cbac4c7 4956 if (name) {
92110913 4957 if (namlen > 0)
1edc1566 4958 mg->mg_ptr = savepvn(name, namlen);
daba3364
NC
4959 else if (namlen == HEf_SVKEY) {
4960 /* Yes, this is casting away const. This is only for the case of
4961 HEf_SVKEY. I think we need to document this abberation of the
4962 constness of the API, rather than making name non-const, as
4963 that change propagating outwards a long way. */
4964 mg->mg_ptr = (char*)SvREFCNT_inc_simple_NN((SV *)name);
4965 } else
92110913 4966 mg->mg_ptr = (char *) name;
9cbac4c7 4967 }
53d44271 4968 mg->mg_virtual = (MGVTBL *) vtable;
68795e93 4969
92110913
NIS
4970 mg_magical(sv);
4971 if (SvGMAGICAL(sv))
4972 SvFLAGS(sv) &= ~(SVf_IOK|SVf_NOK|SVf_POK);
4973 return mg;
4974}
4975
4976/*
4977=for apidoc sv_magic
1c846c1f 4978
92110913
NIS
4979Adds magic to an SV. First upgrades C<sv> to type C<SVt_PVMG> if necessary,
4980then adds a new magic item of type C<how> to the head of the magic list.
4981
2d8d5d5a
SH
4982See C<sv_magicext> (which C<sv_magic> now calls) for a description of the
4983handling of the C<name> and C<namlen> arguments.
4984
4509d3fb
SB
4985You need to use C<sv_magicext> to add magic to SvREADONLY SVs and also
4986to add more than one instance of the same 'how'.
4987
92110913
NIS
4988=cut
4989*/
4990
4991void
2b021c53
SS
4992Perl_sv_magic(pTHX_ register SV *const sv, SV *const obj, const int how,
4993 const char *const name, const I32 namlen)
68795e93 4994{
97aff369 4995 dVAR;
53d44271 4996 const MGVTBL *vtable;
92110913 4997 MAGIC* mg;
92110913 4998
7918f24d
NC
4999 PERL_ARGS_ASSERT_SV_MAGIC;
5000
f8c7b90f 5001#ifdef PERL_OLD_COPY_ON_WRITE
765f542d
NC
5002 if (SvIsCOW(sv))
5003 sv_force_normal_flags(sv, 0);
5004#endif
92110913 5005 if (SvREADONLY(sv)) {
d8084ca5
DM
5006 if (
5007 /* its okay to attach magic to shared strings; the subsequent
5008 * upgrade to PVMG will unshare the string */
5009 !(SvFAKE(sv) && SvTYPE(sv) < SVt_PVMG)
5010
5011 && IN_PERL_RUNTIME
92110913
NIS
5012 && how != PERL_MAGIC_regex_global
5013 && how != PERL_MAGIC_bm
5014 && how != PERL_MAGIC_fm
5015 && how != PERL_MAGIC_sv
e6469971 5016 && how != PERL_MAGIC_backref
92110913
NIS
5017 )
5018 {
f1f66076 5019 Perl_croak(aTHX_ "%s", PL_no_modify);
92110913
NIS
5020 }
5021 }
5022 if (SvMAGICAL(sv) || (how == PERL_MAGIC_taint && SvTYPE(sv) >= SVt_PVMG)) {
5023 if (SvMAGIC(sv) && (mg = mg_find(sv, how))) {
68795e93
NIS
5024 /* sv_magic() refuses to add a magic of the same 'how' as an
5025 existing one
92110913 5026 */
2a509ed3 5027 if (how == PERL_MAGIC_taint) {
92110913 5028 mg->mg_len |= 1;
2a509ed3
NC
5029 /* Any scalar which already had taint magic on which someone
5030 (erroneously?) did SvIOK_on() or similar will now be
5031 incorrectly sporting public "OK" flags. */
5032 SvFLAGS(sv) &= ~(SVf_IOK|SVf_NOK|SVf_POK);
5033 }
92110913
NIS
5034 return;
5035 }
5036 }
68795e93 5037
79072805 5038 switch (how) {
14befaf4 5039 case PERL_MAGIC_sv:
92110913 5040 vtable = &PL_vtbl_sv;
79072805 5041 break;
14befaf4 5042 case PERL_MAGIC_overload:
92110913 5043 vtable = &PL_vtbl_amagic;
a0d0e21e 5044 break;
14befaf4 5045 case PERL_MAGIC_overload_elem:
92110913 5046 vtable = &PL_vtbl_amagicelem;
a0d0e21e 5047 break;
14befaf4 5048 case PERL_MAGIC_overload_table:
92110913 5049 vtable = &PL_vtbl_ovrld;
a0d0e21e 5050 break;
14befaf4 5051 case PERL_MAGIC_bm:
92110913 5052 vtable = &PL_vtbl_bm;
79072805 5053 break;
14befaf4 5054 case PERL_MAGIC_regdata:
92110913 5055 vtable = &PL_vtbl_regdata;
6cef1e77 5056 break;
14befaf4 5057 case PERL_MAGIC_regdatum:
92110913 5058 vtable = &PL_vtbl_regdatum;
6cef1e77 5059 break;
14befaf4 5060 case PERL_MAGIC_env:
92110913 5061 vtable = &PL_vtbl_env;
79072805 5062 break;
14befaf4 5063 case PERL_MAGIC_fm:
92110913 5064 vtable = &PL_vtbl_fm;
55497cff 5065 break;
14befaf4 5066 case PERL_MAGIC_envelem:
92110913 5067 vtable = &PL_vtbl_envelem;
79072805 5068 break;
14befaf4 5069 case PERL_MAGIC_regex_global:
92110913 5070 vtable = &PL_vtbl_mglob;
93a17b20 5071 break;
14befaf4 5072 case PERL_MAGIC_isa:
92110913 5073 vtable = &PL_vtbl_isa;
463ee0b2 5074 break;
14befaf4 5075 case PERL_MAGIC_isaelem:
92110913 5076 vtable = &PL_vtbl_isaelem;
463ee0b2 5077 break;
14befaf4 5078 case PERL_MAGIC_nkeys:
92110913 5079 vtable = &PL_vtbl_nkeys;
16660edb 5080 break;
14befaf4 5081 case PERL_MAGIC_dbfile:
aec46f14 5082 vtable = NULL;
93a17b20 5083 break;
14befaf4 5084 case PERL_MAGIC_dbline:
92110913 5085 vtable = &PL_vtbl_dbline;
79072805 5086 break;
36477c24 5087#ifdef USE_LOCALE_COLLATE
14befaf4 5088 case PERL_MAGIC_collxfrm:
92110913 5089 vtable = &PL_vtbl_collxfrm;
bbce6d69 5090 break;
36477c24 5091#endif /* USE_LOCALE_COLLATE */
14befaf4 5092 case PERL_MAGIC_tied:
92110913 5093 vtable = &PL_vtbl_pack;
463ee0b2 5094 break;
14befaf4
DM
5095 case PERL_MAGIC_tiedelem:
5096 case PERL_MAGIC_tiedscalar:
92110913 5097 vtable = &PL_vtbl_packelem;
463ee0b2 5098 break;
14befaf4 5099 case PERL_MAGIC_qr:
92110913 5100 vtable = &PL_vtbl_regexp;
c277df42 5101 break;
14befaf4 5102 case PERL_MAGIC_sig:
92110913 5103 vtable = &PL_vtbl_sig;
79072805 5104 break;
14befaf4 5105 case PERL_MAGIC_sigelem:
92110913 5106 vtable = &PL_vtbl_sigelem;
79072805 5107 break;
14befaf4 5108 case PERL_MAGIC_taint:
92110913 5109 vtable = &PL_vtbl_taint;
463ee0b2 5110 break;
14befaf4 5111 case PERL_MAGIC_uvar:
92110913 5112 vtable = &PL_vtbl_uvar;
79072805 5113 break;
14befaf4 5114 case PERL_MAGIC_vec:
92110913 5115 vtable = &PL_vtbl_vec;
79072805 5116 break;
a3874608 5117 case PERL_MAGIC_arylen_p:
bfcb3514 5118 case PERL_MAGIC_rhash:
8d2f4536 5119 case PERL_MAGIC_symtab:
ece467f9 5120 case PERL_MAGIC_vstring:
aec46f14 5121 vtable = NULL;
ece467f9 5122 break;
7e8c5dac
HS
5123 case PERL_MAGIC_utf8:
5124 vtable = &PL_vtbl_utf8;
5125 break;
14befaf4 5126 case PERL_MAGIC_substr:
92110913 5127 vtable = &PL_vtbl_substr;
79072805 5128 break;
14befaf4 5129 case PERL_MAGIC_defelem:
92110913 5130 vtable = &PL_vtbl_defelem;
5f05dabc 5131 break;
14befaf4 5132 case PERL_MAGIC_arylen:
92110913 5133 vtable = &PL_vtbl_arylen;
79072805 5134 break;
14befaf4 5135 case PERL_MAGIC_pos:
92110913 5136 vtable = &PL_vtbl_pos;
a0d0e21e 5137 break;
14befaf4 5138 case PERL_MAGIC_backref:
92110913 5139 vtable = &PL_vtbl_backref;
810b8aa5 5140 break;
b3ca2e83
NC
5141 case PERL_MAGIC_hintselem:
5142 vtable = &PL_vtbl_hintselem;
5143 break;
f747ebd6
Z
5144 case PERL_MAGIC_hints:
5145 vtable = &PL_vtbl_hints;
5146 break;
14befaf4
DM
5147 case PERL_MAGIC_ext:
5148 /* Reserved for use by extensions not perl internals. */
4633a7c4
LW
5149 /* Useful for attaching extension internal data to perl vars. */
5150 /* Note that multiple extensions may clash if magical scalars */
5151 /* etc holding private data from one are passed to another. */
aec46f14 5152 vtable = NULL;
a0d0e21e 5153 break;
79072805 5154 default:
14befaf4 5155 Perl_croak(aTHX_ "Don't know how to handle magic of type \\%o", how);
463ee0b2 5156 }
68795e93 5157
92110913 5158 /* Rest of work is done else where */
aec46f14 5159 mg = sv_magicext(sv,obj,how,vtable,name,namlen);
68795e93 5160
92110913
NIS
5161 switch (how) {
5162 case PERL_MAGIC_taint:
5163 mg->mg_len = 1;
5164 break;
5165 case PERL_MAGIC_ext:
5166 case PERL_MAGIC_dbfile:
5167 SvRMAGICAL_on(sv);
5168 break;
5169 }
463ee0b2
LW
5170}
5171
c461cf8f
JH
5172/*
5173=for apidoc sv_unmagic
5174
645c22ef 5175Removes all magic of type C<type> from an SV.
c461cf8f
JH
5176
5177=cut
5178*/
5179
463ee0b2 5180int
2b021c53 5181Perl_sv_unmagic(pTHX_ SV *const sv, const int type)
463ee0b2
LW
5182{
5183 MAGIC* mg;
5184 MAGIC** mgp;
7918f24d
NC
5185
5186 PERL_ARGS_ASSERT_SV_UNMAGIC;
5187
91bba347 5188 if (SvTYPE(sv) < SVt_PVMG || !SvMAGIC(sv))
463ee0b2 5189 return 0;
064cf529 5190 mgp = &(((XPVMG*) SvANY(sv))->xmg_u.xmg_magic);
463ee0b2
LW
5191 for (mg = *mgp; mg; mg = *mgp) {
5192 if (mg->mg_type == type) {
e1ec3a88 5193 const MGVTBL* const vtbl = mg->mg_virtual;
463ee0b2 5194 *mgp = mg->mg_moremagic;
1d7c1841 5195 if (vtbl && vtbl->svt_free)
fc0dc3b3 5196 CALL_FPTR(vtbl->svt_free)(aTHX_ sv, mg);
14befaf4 5197 if (mg->mg_ptr && mg->mg_type != PERL_MAGIC_regex_global) {
92110913 5198 if (mg->mg_len > 0)
1edc1566 5199 Safefree(mg->mg_ptr);
565764a8 5200 else if (mg->mg_len == HEf_SVKEY)
daba3364 5201 SvREFCNT_dec(MUTABLE_SV(mg->mg_ptr));
d2923cdd 5202 else if (mg->mg_type == PERL_MAGIC_utf8)
7e8c5dac 5203 Safefree(mg->mg_ptr);
9cbac4c7 5204 }
a0d0e21e
LW
5205 if (mg->mg_flags & MGf_REFCOUNTED)
5206 SvREFCNT_dec(mg->mg_obj);
463ee0b2
LW
5207 Safefree(mg);
5208 }
5209 else
5210 mgp = &mg->mg_moremagic;
79072805 5211 }
806e7ca7
CS
5212 if (SvMAGIC(sv)) {
5213 if (SvMAGICAL(sv)) /* if we're under save_magic, wait for restore_magic; */
5214 mg_magical(sv); /* else fix the flags now */
5215 }
5216 else {
463ee0b2 5217 SvMAGICAL_off(sv);
c268c2a6 5218 SvFLAGS(sv) |= (SvFLAGS(sv) & (SVp_IOK|SVp_NOK|SVp_POK)) >> PRIVSHIFT;
463ee0b2 5219 }
463ee0b2 5220 return 0;
79072805
LW
5221}
5222
c461cf8f
JH
5223/*
5224=for apidoc sv_rvweaken
5225
645c22ef
DM
5226Weaken a reference: set the C<SvWEAKREF> flag on this RV; give the
5227referred-to SV C<PERL_MAGIC_backref> magic if it hasn't already; and
5228push a back-reference to this RV onto the array of backreferences
1e73acc8
AS
5229associated with that magic. If the RV is magical, set magic will be
5230called after the RV is cleared.
c461cf8f
JH
5231
5232=cut
5233*/
5234
810b8aa5 5235SV *
2b021c53 5236Perl_sv_rvweaken(pTHX_ SV *const sv)
810b8aa5
GS
5237{
5238 SV *tsv;
7918f24d
NC
5239
5240 PERL_ARGS_ASSERT_SV_RVWEAKEN;
5241
810b8aa5
GS
5242 if (!SvOK(sv)) /* let undefs pass */
5243 return sv;
5244 if (!SvROK(sv))
cea2e8a9 5245 Perl_croak(aTHX_ "Can't weaken a nonreference");
810b8aa5 5246 else if (SvWEAKREF(sv)) {
a2a5de95 5247 Perl_ck_warner(aTHX_ packWARN(WARN_MISC), "Reference is already weak");
810b8aa5
GS
5248 return sv;
5249 }
5250 tsv = SvRV(sv);
e15faf7d 5251 Perl_sv_add_backref(aTHX_ tsv, sv);
810b8aa5 5252 SvWEAKREF_on(sv);
1c846c1f 5253 SvREFCNT_dec(tsv);
810b8aa5
GS
5254 return sv;
5255}
5256
645c22ef
DM
5257/* Give tsv backref magic if it hasn't already got it, then push a
5258 * back-reference to sv onto the array associated with the backref magic.
5259 */
5260
fd996479
DM
5261/* A discussion about the backreferences array and its refcount:
5262 *
5263 * The AV holding the backreferences is pointed to either as the mg_obj of
5264 * PERL_MAGIC_backref, or in the specific case of a HV that has the hv_aux
5265 * structure, from the xhv_backreferences field. (A HV without hv_aux will
5266 * have the standard magic instead.) The array is created with a refcount
5267 * of 2. This means that if during global destruction the array gets
5268 * picked on first to have its refcount decremented by the random zapper,
5269 * it won't actually be freed, meaning it's still theere for when its
5270 * parent gets freed.
5271 * When the parent SV is freed, in the case of magic, the magic is freed,
5272 * Perl_magic_killbackrefs is called which decrements one refcount, then
5273 * mg_obj is freed which kills the second count.
5274 * In the vase of a HV being freed, one ref is removed by
5275 * Perl_hv_kill_backrefs, the other by Perl_sv_kill_backrefs, which it
5276 * calls.
5277 */
5278
e15faf7d 5279void
2b021c53 5280Perl_sv_add_backref(pTHX_ SV *const tsv, SV *const sv)
810b8aa5 5281{
97aff369 5282 dVAR;
810b8aa5 5283 AV *av;
86f55936 5284
7918f24d
NC
5285 PERL_ARGS_ASSERT_SV_ADD_BACKREF;
5286
86f55936 5287 if (SvTYPE(tsv) == SVt_PVHV) {
85fbaab2 5288 AV **const avp = Perl_hv_backreferences_p(aTHX_ MUTABLE_HV(tsv));
86f55936
NC
5289
5290 av = *avp;
5291 if (!av) {
5292 /* There is no AV in the offical place - try a fixup. */
5293 MAGIC *const mg = mg_find(tsv, PERL_MAGIC_backref);
5294
5295 if (mg) {
5296 /* Aha. They've got it stowed in magic. Bring it back. */
502c6561 5297 av = MUTABLE_AV(mg->mg_obj);
86f55936
NC
5298 /* Stop mg_free decreasing the refernce count. */
5299 mg->mg_obj = NULL;
5300 /* Stop mg_free even calling the destructor, given that
5301 there's no AV to free up. */
5302 mg->mg_virtual = 0;
5303 sv_unmagic(tsv, PERL_MAGIC_backref);
5304 } else {
5305 av = newAV();
5306 AvREAL_off(av);
fd996479 5307 SvREFCNT_inc_simple_void(av); /* see discussion above */
86f55936
NC
5308 }
5309 *avp = av;
5310 }
5311 } else {
5312 const MAGIC *const mg
5313 = SvMAGICAL(tsv) ? mg_find(tsv, PERL_MAGIC_backref) : NULL;
5314 if (mg)
502c6561 5315 av = MUTABLE_AV(mg->mg_obj);
86f55936
NC
5316 else {
5317 av = newAV();
5318 AvREAL_off(av);
daba3364 5319 sv_magic(tsv, MUTABLE_SV(av), PERL_MAGIC_backref, NULL, 0);
fd996479 5320 /* av now has a refcnt of 2; see discussion above */
86f55936 5321 }
810b8aa5 5322 }
d91d49e8 5323 if (AvFILLp(av) >= AvMAX(av)) {
d91d49e8
MM
5324 av_extend(av, AvFILLp(av)+1);
5325 }
5326 AvARRAY(av)[++AvFILLp(av)] = sv; /* av_push() */
810b8aa5
GS
5327}
5328
645c22ef
DM
5329/* delete a back-reference to ourselves from the backref magic associated
5330 * with the SV we point to.
5331 */
5332
1c846c1f 5333STATIC void
2b021c53 5334S_sv_del_backref(pTHX_ SV *const tsv, SV *const sv)
810b8aa5 5335{
97aff369 5336 dVAR;
86f55936 5337 AV *av = NULL;
810b8aa5
GS
5338 SV **svp;
5339 I32 i;
86f55936 5340
7918f24d
NC
5341 PERL_ARGS_ASSERT_SV_DEL_BACKREF;
5342
86f55936 5343 if (SvTYPE(tsv) == SVt_PVHV && SvOOK(tsv)) {
85fbaab2 5344 av = *Perl_hv_backreferences_p(aTHX_ MUTABLE_HV(tsv));
5b285ea4
NC
5345 /* We mustn't attempt to "fix up" the hash here by moving the
5346 backreference array back to the hv_aux structure, as that is stored
5347 in the main HvARRAY(), and hfreentries assumes that no-one
5348 reallocates HvARRAY() while it is running. */
86f55936
NC
5349 }
5350 if (!av) {
5351 const MAGIC *const mg
5352 = SvMAGICAL(tsv) ? mg_find(tsv, PERL_MAGIC_backref) : NULL;
5353 if (mg)
502c6561 5354 av = MUTABLE_AV(mg->mg_obj);
86f55936 5355 }
41fae7a1
DM
5356
5357 if (!av)
cea2e8a9 5358 Perl_croak(aTHX_ "panic: del_backref");
86f55936 5359
41fae7a1 5360 assert(!SvIS_FREED(av));
86f55936 5361
810b8aa5 5362 svp = AvARRAY(av);
6a76db8b
NC
5363 /* We shouldn't be in here more than once, but for paranoia reasons lets
5364 not assume this. */
5365 for (i = AvFILLp(av); i >= 0; i--) {
5366 if (svp[i] == sv) {
5367 const SSize_t fill = AvFILLp(av);
5368 if (i != fill) {
5369 /* We weren't the last entry.
5370 An unordered list has this property that you can take the
5371 last element off the end to fill the hole, and it's still
5372 an unordered list :-)
5373 */
5374 svp[i] = svp[fill];
5375 }
a0714e2c 5376 svp[fill] = NULL;
6a76db8b
NC
5377 AvFILLp(av) = fill - 1;
5378 }
5379 }
810b8aa5
GS
5380}
5381
86f55936 5382int
2b021c53 5383Perl_sv_kill_backrefs(pTHX_ SV *const sv, AV *const av)
86f55936
NC
5384{
5385 SV **svp = AvARRAY(av);
5386
7918f24d 5387 PERL_ARGS_ASSERT_SV_KILL_BACKREFS;
86f55936
NC
5388 PERL_UNUSED_ARG(sv);
5389
41fae7a1
DM
5390 assert(!svp || !SvIS_FREED(av));
5391 if (svp) {
86f55936
NC
5392 SV *const *const last = svp + AvFILLp(av);
5393
5394 while (svp <= last) {
5395 if (*svp) {
5396 SV *const referrer = *svp;
5397 if (SvWEAKREF(referrer)) {
5398 /* XXX Should we check that it hasn't changed? */
5399 SvRV_set(referrer, 0);
5400 SvOK_off(referrer);
5401 SvWEAKREF_off(referrer);
1e73acc8 5402 SvSETMAGIC(referrer);
86f55936
NC
5403 } else if (SvTYPE(referrer) == SVt_PVGV ||
5404 SvTYPE(referrer) == SVt_PVLV) {
5405 /* You lookin' at me? */
5406 assert(GvSTASH(referrer));
1d193675 5407 assert(GvSTASH(referrer) == (const HV *)sv);
86f55936
NC
5408 GvSTASH(referrer) = 0;
5409 } else {
5410 Perl_croak(aTHX_
5411 "panic: magic_killbackrefs (flags=%"UVxf")",
5412 (UV)SvFLAGS(referrer));
5413 }
5414
a0714e2c 5415 *svp = NULL;
86f55936
NC
5416 }
5417 svp++;
5418 }
5419 }
5420 SvREFCNT_dec(av); /* remove extra count added by sv_add_backref() */
5421 return 0;
5422}
5423
954c1994
GS
5424/*
5425=for apidoc sv_insert
5426
5427Inserts a string at the specified offset/length within the SV. Similar to
c0dd94a0 5428the Perl substr() function. Handles get magic.
954c1994 5429
c0dd94a0
VP
5430=for apidoc sv_insert_flags
5431
5432Same as C<sv_insert>, but the extra C<flags> are passed the C<SvPV_force_flags> that applies to C<bigstr>.
5433
5434=cut
5435*/
5436
5437void
5438Perl_sv_insert_flags(pTHX_ SV *const bigstr, const STRLEN offset, const STRLEN len, const char *const little, const STRLEN littlelen, const U32 flags)
5439{
97aff369 5440 dVAR;
79072805
LW
5441 register char *big;
5442 register char *mid;
5443 register char *midend;
5444 register char *bigend;
5445 register I32 i;
6ff81951 5446 STRLEN curlen;
1c846c1f 5447
27aecdc6 5448 PERL_ARGS_ASSERT_SV_INSERT_FLAGS;
79072805 5449
8990e307 5450 if (!bigstr)
cea2e8a9 5451 Perl_croak(aTHX_ "Can't modify non-existent substring");
c0dd94a0 5452 SvPV_force_flags(bigstr, curlen, flags);
60fa28ff 5453 (void)SvPOK_only_UTF8(bigstr);
6ff81951
GS
5454 if (offset + len > curlen) {
5455 SvGROW(bigstr, offset+len+1);
93524f2b 5456 Zero(SvPVX(bigstr)+curlen, offset+len-curlen, char);
6ff81951
GS
5457 SvCUR_set(bigstr, offset+len);
5458 }
79072805 5459
69b47968 5460 SvTAINT(bigstr);
79072805
LW
5461 i = littlelen - len;
5462 if (i > 0) { /* string might grow */
a0d0e21e 5463 big = SvGROW(bigstr, SvCUR(bigstr) + i + 1);
79072805
LW
5464 mid = big + offset + len;
5465 midend = bigend = big + SvCUR(bigstr);
5466 bigend += i;
5467 *bigend = '\0';
5468 while (midend > mid) /* shove everything down */
5469 *--bigend = *--midend;
5470 Move(little,big+offset,littlelen,char);
b162af07 5471 SvCUR_set(bigstr, SvCUR(bigstr) + i);
79072805
LW
5472 SvSETMAGIC(bigstr);
5473 return;
5474 }
5475 else if (i == 0) {
463ee0b2 5476 Move(little,SvPVX(bigstr)+offset,len,char);
79072805
LW
5477 SvSETMAGIC(bigstr);
5478 return;
5479 }
5480
463ee0b2 5481 big = SvPVX(bigstr);
79072805
LW
5482 mid = big + offset;
5483 midend = mid + len;
5484 bigend = big + SvCUR(bigstr);
5485
5486 if (midend > bigend)
cea2e8a9 5487 Perl_croak(aTHX_ "panic: sv_insert");
79072805
LW
5488
5489 if (mid - big > bigend - midend) { /* faster to shorten from end */
5490 if (littlelen) {
5491 Move(little, mid, littlelen,char);
5492 mid += littlelen;
5493 }
5494 i = bigend - midend;
5495 if (i > 0) {
5496 Move(midend, mid, i,char);
5497 mid += i;
5498 }
5499 *mid = '\0';
5500 SvCUR_set(bigstr, mid - big);
5501 }
155aba94 5502 else if ((i = mid - big)) { /* faster from front */
79072805
LW
5503 midend -= littlelen;
5504 mid = midend;
0d3c21b0 5505 Move(big, midend - i, i, char);
79072805 5506 sv_chop(bigstr,midend-i);
79072805
LW
5507 if (littlelen)
5508 Move(little, mid, littlelen,char);
5509 }
5510 else if (littlelen) {
5511 midend -= littlelen;
5512 sv_chop(bigstr,midend);
5513 Move(little,midend,littlelen,char);
5514 }
5515 else {
5516 sv_chop(bigstr,midend);
5517 }
5518 SvSETMAGIC(bigstr);
5519}
5520
c461cf8f
JH
5521/*
5522=for apidoc sv_replace
5523
5524Make the first argument a copy of the second, then delete the original.
645c22ef
DM
5525The target SV physically takes over ownership of the body of the source SV
5526and inherits its flags; however, the target keeps any magic it owns,
5527and any magic in the source is discarded.
ff276b08 5528Note that this is a rather specialist SV copying operation; most of the
645c22ef 5529time you'll want to use C<sv_setsv> or one of its many macro front-ends.
c461cf8f
JH
5530
5531=cut
5532*/
79072805
LW
5533
5534void
af828c01 5535Perl_sv_replace(pTHX_ register SV *const sv, register SV *const nsv)
79072805 5536{
97aff369 5537 dVAR;
a3b680e6 5538 const U32 refcnt = SvREFCNT(sv);
7918f24d
NC
5539
5540 PERL_ARGS_ASSERT_SV_REPLACE;
5541
765f542d 5542 SV_CHECK_THINKFIRST_COW_DROP(sv);
30e5c352 5543 if (SvREFCNT(nsv) != 1) {
fe13d51d
JM
5544 Perl_croak(aTHX_ "panic: reference miscount on nsv in sv_replace()"
5545 " (%" UVuf " != 1)", (UV) SvREFCNT(nsv));
30e5c352 5546 }
93a17b20 5547 if (SvMAGICAL(sv)) {
a0d0e21e
LW
5548 if (SvMAGICAL(nsv))
5549 mg_free(nsv);
5550 else
5551 sv_upgrade(nsv, SVt_PVMG);
b162af07 5552 SvMAGIC_set(nsv, SvMAGIC(sv));
a0d0e21e 5553 SvFLAGS(nsv) |= SvMAGICAL(sv);
93a17b20 5554 SvMAGICAL_off(sv);
b162af07 5555 SvMAGIC_set(sv, NULL);
93a17b20 5556 }
79072805
LW
5557 SvREFCNT(sv) = 0;
5558 sv_clear(sv);
477f5d66 5559 assert(!SvREFCNT(sv));
fd0854ff
DM
5560#ifdef DEBUG_LEAKING_SCALARS
5561 sv->sv_flags = nsv->sv_flags;
5562 sv->sv_any = nsv->sv_any;
5563 sv->sv_refcnt = nsv->sv_refcnt;
f34d0642 5564 sv->sv_u = nsv->sv_u;
fd0854ff 5565#else
79072805 5566 StructCopy(nsv,sv,SV);
fd0854ff 5567#endif
4df7f6af 5568 if(SvTYPE(sv) == SVt_IV) {
7b2c381c 5569 SvANY(sv)
339049b0 5570 = (XPVIV*)((char*)&(sv->sv_u.svu_iv) - STRUCT_OFFSET(XPVIV, xiv_iv));
7b2c381c
NC
5571 }
5572
fd0854ff 5573
f8c7b90f 5574#ifdef PERL_OLD_COPY_ON_WRITE
d3d0e6f1
NC
5575 if (SvIsCOW_normal(nsv)) {
5576 /* We need to follow the pointers around the loop to make the
5577 previous SV point to sv, rather than nsv. */
5578 SV *next;
5579 SV *current = nsv;
5580 while ((next = SV_COW_NEXT_SV(current)) != nsv) {
5581 assert(next);
5582 current = next;
3f7c398e 5583 assert(SvPVX_const(current) == SvPVX_const(nsv));
d3d0e6f1
NC
5584 }
5585 /* Make the SV before us point to the SV after us. */
5586 if (DEBUG_C_TEST) {
5587 PerlIO_printf(Perl_debug_log, "previous is\n");
5588 sv_dump(current);
a29f6d03
NC
5589 PerlIO_printf(Perl_debug_log,
5590 "move it from 0x%"UVxf" to 0x%"UVxf"\n",
d3d0e6f1
NC
5591 (UV) SV_COW_NEXT_SV(current), (UV) sv);
5592 }
a29f6d03 5593 SV_COW_NEXT_SV_SET(current, sv);
d3d0e6f1
NC
5594 }
5595#endif
79072805 5596 SvREFCNT(sv) = refcnt;
1edc1566 5597 SvFLAGS(nsv) |= SVTYPEMASK; /* Mark as freed */
39cf41c2 5598 SvREFCNT(nsv) = 0;
463ee0b2 5599 del_SV(nsv);
79072805
LW
5600}
5601
c461cf8f
JH
5602/*
5603=for apidoc sv_clear
5604
645c22ef
DM
5605Clear an SV: call any destructors, free up any memory used by the body,
5606and free the body itself. The SV's head is I<not> freed, although
5607its type is set to all 1's so that it won't inadvertently be assumed
5608to be live during global destruction etc.
5609This function should only be called when REFCNT is zero. Most of the time
5610you'll want to call C<sv_free()> (or its macro wrapper C<SvREFCNT_dec>)
5611instead.
c461cf8f
JH
5612
5613=cut
5614*/
5615
79072805 5616void
af828c01 5617Perl_sv_clear(pTHX_ register SV *const sv)
79072805 5618{
27da23d5 5619 dVAR;
82bb6deb 5620 const U32 type = SvTYPE(sv);
8edfc514
NC
5621 const struct body_details *const sv_type_details
5622 = bodies_by_type + type;
dd69841b 5623 HV *stash;
82bb6deb 5624
7918f24d 5625 PERL_ARGS_ASSERT_SV_CLEAR;
79072805 5626 assert(SvREFCNT(sv) == 0);
ceb531cd 5627 assert(SvTYPE(sv) != SVTYPEMASK);
79072805 5628
d2a0f284
JC
5629 if (type <= SVt_IV) {
5630 /* See the comment in sv.h about the collusion between this early
5631 return and the overloading of the NULL and IV slots in the size
5632 table. */
4df7f6af
NC
5633 if (SvROK(sv)) {
5634 SV * const target = SvRV(sv);
5635 if (SvWEAKREF(sv))
5636 sv_del_backref(target, sv);
5637 else
5638 SvREFCNT_dec(target);
5639 }
5640 SvFLAGS(sv) &= SVf_BREAK;
5641 SvFLAGS(sv) |= SVTYPEMASK;
82bb6deb 5642 return;
d2a0f284 5643 }
82bb6deb 5644
ed6116ce 5645 if (SvOBJECT(sv)) {
eba16661
JH
5646 if (PL_defstash && /* Still have a symbol table? */
5647 SvDESTROYABLE(sv))
5648 {
39644a26 5649 dSP;
893645bd 5650 HV* stash;
d460ef45 5651 do {
b464bac0 5652 CV* destructor;
4e8e7886 5653 stash = SvSTASH(sv);
32251b26 5654 destructor = StashHANDLER(stash,DESTROY);
fbb3ee5a 5655 if (destructor
99ab892b
NC
5656 /* A constant subroutine can have no side effects, so
5657 don't bother calling it. */
5658 && !CvCONST(destructor)
fbb3ee5a
RGS
5659 /* Don't bother calling an empty destructor */
5660 && (CvISXSUB(destructor)
5661 || CvSTART(destructor)->op_next->op_type != OP_LEAVESUB))
5662 {
1b6737cc 5663 SV* const tmpref = newRV(sv);
5cc433a6 5664 SvREADONLY_on(tmpref); /* DESTROY() could be naughty */
4e8e7886 5665 ENTER;
e788e7d3 5666 PUSHSTACKi(PERLSI_DESTROY);
4e8e7886
GS
5667 EXTEND(SP, 2);
5668 PUSHMARK(SP);
5cc433a6 5669 PUSHs(tmpref);
4e8e7886 5670 PUTBACK;
daba3364 5671 call_sv(MUTABLE_SV(destructor), G_DISCARD|G_EVAL|G_KEEPERR|G_VOID);
7a5fa8a2
NIS
5672
5673
d3acc0f7 5674 POPSTACK;
3095d977 5675 SPAGAIN;
4e8e7886 5676 LEAVE;
5cc433a6
AB
5677 if(SvREFCNT(tmpref) < 2) {
5678 /* tmpref is not kept alive! */
5679 SvREFCNT(sv)--;
b162af07 5680 SvRV_set(tmpref, NULL);
5cc433a6
AB
5681 SvROK_off(tmpref);
5682 }
5683 SvREFCNT_dec(tmpref);
4e8e7886
GS
5684 }
5685 } while (SvOBJECT(sv) && SvSTASH(sv) != stash);
8ebc5c01 5686
6f44e0a4
JP
5687
5688 if (SvREFCNT(sv)) {
5689 if (PL_in_clean_objs)
cea2e8a9 5690 Perl_croak(aTHX_ "DESTROY created new reference to dead object '%s'",
bfcb3514 5691 HvNAME_get(stash));
6f44e0a4
JP
5692 /* DESTROY gave object new lease on life */
5693 return;
5694 }
a0d0e21e 5695 }
4e8e7886 5696
a0d0e21e 5697 if (SvOBJECT(sv)) {
4e8e7886 5698 SvREFCNT_dec(SvSTASH(sv)); /* possibly of changed persuasion */
a0d0e21e 5699 SvOBJECT_off(sv); /* Curse the object. */
82bb6deb 5700 if (type != SVt_PVIO)
3280af22 5701 --PL_sv_objcount; /* XXX Might want something more general */
a0d0e21e 5702 }
463ee0b2 5703 }
82bb6deb 5704 if (type >= SVt_PVMG) {
cecf5685 5705 if (type == SVt_PVMG && SvPAD_OUR(sv)) {
73d95100 5706 SvREFCNT_dec(SvOURSTASH(sv));
e736a858 5707 } else if (SvMAGIC(sv))
524189f1 5708 mg_free(sv);
00b1698f 5709 if (type == SVt_PVMG && SvPAD_TYPED(sv))
524189f1
JH
5710 SvREFCNT_dec(SvSTASH(sv));
5711 }
82bb6deb 5712 switch (type) {
cecf5685 5713 /* case SVt_BIND: */
8990e307 5714 case SVt_PVIO:
df0bd2f4
GS
5715 if (IoIFP(sv) &&
5716 IoIFP(sv) != PerlIO_stdin() &&
5f05dabc 5717 IoIFP(sv) != PerlIO_stdout() &&
5718 IoIFP(sv) != PerlIO_stderr())
93578b34 5719 {
a45c7426 5720 io_close(MUTABLE_IO(sv), FALSE);
93578b34 5721 }
1d7c1841 5722 if (IoDIRP(sv) && !(IoFLAGS(sv) & IOf_FAKE_DIRP))
1236053a 5723 PerlDir_close(IoDIRP(sv));
1d7c1841 5724 IoDIRP(sv) = (DIR*)NULL;
8990e307
LW
5725 Safefree(IoTOP_NAME(sv));
5726 Safefree(IoFMT_NAME(sv));
5727 Safefree(IoBOTTOM_NAME(sv));
82bb6deb 5728 goto freescalar;
5c35adbb 5729 case SVt_REGEXP:
288b8c02 5730 /* FIXME for plugins */
d2f13c59 5731 pregfree2((REGEXP*) sv);
5c35adbb 5732 goto freescalar;
79072805 5733 case SVt_PVCV:
748a9306 5734 case SVt_PVFM:
ea726b52 5735 cv_undef(MUTABLE_CV(sv));
a0d0e21e 5736 goto freescalar;
79072805 5737 case SVt_PVHV:
1d193675 5738 if (PL_last_swash_hv == (const HV *)sv) {
e7fab884
NC
5739 PL_last_swash_hv = NULL;
5740 }
85fbaab2
NC
5741 Perl_hv_kill_backrefs(aTHX_ MUTABLE_HV(sv));
5742 hv_undef(MUTABLE_HV(sv));
a0d0e21e 5743 break;
79072805 5744 case SVt_PVAV:
502c6561 5745 if (PL_comppad == MUTABLE_AV(sv)) {
3f90d085
DM
5746 PL_comppad = NULL;
5747 PL_curpad = NULL;
5748 }
502c6561 5749 av_undef(MUTABLE_AV(sv));
a0d0e21e 5750 break;
02270b4e 5751 case SVt_PVLV:
dd28f7bb
DM
5752 if (LvTYPE(sv) == 'T') { /* for tie: return HE to pool */
5753 SvREFCNT_dec(HeKEY_sv((HE*)LvTARG(sv)));
5754 HeNEXT((HE*)LvTARG(sv)) = PL_hv_fetch_ent_mh;
5755 PL_hv_fetch_ent_mh = (HE*)LvTARG(sv);
5756 }
5757 else if (LvTYPE(sv) != 't') /* unless tie: unrefcnted fake SV** */
5758 SvREFCNT_dec(LvTARG(sv));
a0d0e21e 5759 case SVt_PVGV:
cecf5685 5760 if (isGV_with_GP(sv)) {
159b6efe
NC
5761 if(GvCVu((const GV *)sv) && (stash = GvSTASH(MUTABLE_GV(sv)))
5762 && HvNAME_get(stash))
dd69841b 5763 mro_method_changed_in(stash);
159b6efe 5764 gp_free(MUTABLE_GV(sv));
cecf5685
NC
5765 if (GvNAME_HEK(sv))
5766 unshare_hek(GvNAME_HEK(sv));
dd69841b
BB
5767 /* If we're in a stash, we don't own a reference to it. However it does
5768 have a back reference to us, which needs to be cleared. */
5769 if (!SvVALID(sv) && (stash = GvSTASH(sv)))
daba3364 5770 sv_del_backref(MUTABLE_SV(stash), sv);
cecf5685 5771 }
8571fe2f
NC
5772 /* FIXME. There are probably more unreferenced pointers to SVs in the
5773 interpreter struct that we should check and tidy in a similar
5774 fashion to this: */
159b6efe 5775 if ((const GV *)sv == PL_last_in_gv)
8571fe2f 5776 PL_last_in_gv = NULL;
79072805 5777 case SVt_PVMG:
79072805
LW
5778 case SVt_PVNV:
5779 case SVt_PVIV:
7a4bba22 5780 case SVt_PV:
a0d0e21e 5781 freescalar:
5228ca4e
NC
5782 /* Don't bother with SvOOK_off(sv); as we're only going to free it. */
5783 if (SvOOK(sv)) {
69240efd
NC
5784 STRLEN offset;
5785 SvOOK_offset(sv, offset);
5786 SvPV_set(sv, SvPVX_mutable(sv) - offset);
5228ca4e
NC
5787 /* Don't even bother with turning off the OOK flag. */
5788 }
810b8aa5 5789 if (SvROK(sv)) {
b37c2d43 5790 SV * const target = SvRV(sv);
810b8aa5 5791 if (SvWEAKREF(sv))
e15faf7d 5792 sv_del_backref(target, sv);
810b8aa5 5793 else
e15faf7d 5794 SvREFCNT_dec(target);
810b8aa5 5795 }
f8c7b90f 5796#ifdef PERL_OLD_COPY_ON_WRITE
3f7c398e 5797 else if (SvPVX_const(sv)) {
765f542d 5798 if (SvIsCOW(sv)) {
46187eeb
NC
5799 if (DEBUG_C_TEST) {
5800 PerlIO_printf(Perl_debug_log, "Copy on write: clear\n");
e419cbc5 5801 sv_dump(sv);
46187eeb 5802 }
5302ffd4
NC
5803 if (SvLEN(sv)) {
5804 sv_release_COW(sv, SvPVX_const(sv), SV_COW_NEXT_SV(sv));
5805 } else {
5806 unshare_hek(SvSHARED_HEK_FROM_PV(SvPVX_const(sv)));
5807 }
5808
765f542d
NC
5809 SvFAKE_off(sv);
5810 } else if (SvLEN(sv)) {
3f7c398e 5811 Safefree(SvPVX_const(sv));
765f542d
NC
5812 }
5813 }
5814#else
3f7c398e 5815 else if (SvPVX_const(sv) && SvLEN(sv))
94010e71 5816 Safefree(SvPVX_mutable(sv));
3f7c398e 5817 else if (SvPVX_const(sv) && SvREADONLY(sv) && SvFAKE(sv)) {
bdd68bc3 5818 unshare_hek(SvSHARED_HEK_FROM_PV(SvPVX_const(sv)));
1c846c1f
NIS
5819 SvFAKE_off(sv);
5820 }
765f542d 5821#endif
79072805
LW
5822 break;
5823 case SVt_NV:
79072805
LW
5824 break;
5825 }
5826
893645bd
NC
5827 SvFLAGS(sv) &= SVf_BREAK;
5828 SvFLAGS(sv) |= SVTYPEMASK;
5829
8edfc514 5830 if (sv_type_details->arena) {
b9502f15 5831 del_body(((char *)SvANY(sv) + sv_type_details->offset),
8edfc514
NC
5832 &PL_body_roots[type]);
5833 }
d2a0f284 5834 else if (sv_type_details->body_size) {
8edfc514
NC
5835 my_safefree(SvANY(sv));
5836 }
79072805
LW
5837}
5838
645c22ef
DM
5839/*
5840=for apidoc sv_newref
5841
5842Increment an SV's reference count. Use the C<SvREFCNT_inc()> wrapper
5843instead.
5844
5845=cut
5846*/
5847
79072805 5848SV *
af828c01 5849Perl_sv_newref(pTHX_ SV *const sv)
79072805 5850{
96a5add6 5851 PERL_UNUSED_CONTEXT;
463ee0b2 5852 if (sv)
4db098f4 5853 (SvREFCNT(sv))++;
79072805
LW
5854 return sv;
5855}
5856
c461cf8f
JH
5857/*
5858=for apidoc sv_free
5859
645c22ef
DM
5860Decrement an SV's reference count, and if it drops to zero, call
5861C<sv_clear> to invoke destructors and free up any memory used by
5862the body; finally, deallocate the SV's head itself.
5863Normally called via a wrapper macro C<SvREFCNT_dec>.
c461cf8f
JH
5864
5865=cut
5866*/
5867
79072805 5868void
af828c01 5869Perl_sv_free(pTHX_ SV *const sv)
79072805 5870{
27da23d5 5871 dVAR;
79072805
LW
5872 if (!sv)
5873 return;
a0d0e21e
LW
5874 if (SvREFCNT(sv) == 0) {
5875 if (SvFLAGS(sv) & SVf_BREAK)
645c22ef
DM
5876 /* this SV's refcnt has been artificially decremented to
5877 * trigger cleanup */
a0d0e21e 5878 return;
3280af22 5879 if (PL_in_clean_all) /* All is fair */
1edc1566 5880 return;
d689ffdd
JP
5881 if (SvREADONLY(sv) && SvIMMORTAL(sv)) {
5882 /* make sure SvREFCNT(sv)==0 happens very seldom */
5883 SvREFCNT(sv) = (~(U32)0)/2;
5884 return;
5885 }
41e4abd8 5886 if (ckWARN_d(WARN_INTERNAL)) {
41e4abd8
NC
5887#ifdef DEBUG_LEAKING_SCALARS_FORK_DUMP
5888 Perl_dump_sv_child(aTHX_ sv);
e4c5322d
DM
5889#else
5890 #ifdef DEBUG_LEAKING_SCALARS
bfd95973 5891 sv_dump(sv);
e4c5322d 5892 #endif
bfd95973
NC
5893#ifdef DEBUG_LEAKING_SCALARS_ABORT
5894 if (PL_warnhook == PERL_WARNHOOK_FATAL
5895 || ckDEAD(packWARN(WARN_INTERNAL))) {
5896 /* Don't let Perl_warner cause us to escape our fate: */
5897 abort();
5898 }
5899#endif
5900 /* This may not return: */
5901 Perl_warner(aTHX_ packWARN(WARN_INTERNAL),
5902 "Attempt to free unreferenced scalar: SV 0x%"UVxf
5903 pTHX__FORMAT, PTR2UV(sv) pTHX__VALUE);
41e4abd8
NC
5904#endif
5905 }
77abb4c6
NC
5906#ifdef DEBUG_LEAKING_SCALARS_ABORT
5907 abort();
5908#endif
79072805
LW
5909 return;
5910 }
4db098f4 5911 if (--(SvREFCNT(sv)) > 0)
8990e307 5912 return;
8c4d3c90
NC
5913 Perl_sv_free2(aTHX_ sv);
5914}
5915
5916void
af828c01 5917Perl_sv_free2(pTHX_ SV *const sv)
8c4d3c90 5918{
27da23d5 5919 dVAR;
7918f24d
NC
5920
5921 PERL_ARGS_ASSERT_SV_FREE2;
5922
463ee0b2
LW
5923#ifdef DEBUGGING
5924 if (SvTEMP(sv)) {
9b387841
NC
5925 Perl_ck_warner_d(aTHX_ packWARN(WARN_DEBUGGING),
5926 "Attempt to free temp prematurely: SV 0x%"UVxf
5927 pTHX__FORMAT, PTR2UV(sv) pTHX__VALUE);
79072805 5928 return;
79072805 5929 }
463ee0b2 5930#endif
d689ffdd
JP
5931 if (SvREADONLY(sv) && SvIMMORTAL(sv)) {
5932 /* make sure SvREFCNT(sv)==0 happens very seldom */
5933 SvREFCNT(sv) = (~(U32)0)/2;
5934 return;
5935 }
79072805 5936 sv_clear(sv);
477f5d66
CS
5937 if (! SvREFCNT(sv))
5938 del_SV(sv);
79072805
LW
5939}
5940
954c1994
GS
5941/*
5942=for apidoc sv_len
5943
645c22ef
DM
5944Returns the length of the string in the SV. Handles magic and type
5945coercion. See also C<SvCUR>, which gives raw access to the xpv_cur slot.
954c1994
GS
5946
5947=cut
5948*/
5949
79072805 5950STRLEN
af828c01 5951Perl_sv_len(pTHX_ register SV *const sv)
79072805 5952{
463ee0b2 5953 STRLEN len;
79072805
LW
5954
5955 if (!sv)
5956 return 0;
5957
8990e307 5958 if (SvGMAGICAL(sv))
565764a8 5959 len = mg_length(sv);
8990e307 5960 else
4d84ee25 5961 (void)SvPV_const(sv, len);
463ee0b2 5962 return len;
79072805
LW
5963}
5964
c461cf8f
JH
5965/*
5966=for apidoc sv_len_utf8
5967
5968Returns the number of characters in the string in an SV, counting wide
1e54db1a 5969UTF-8 bytes as a single character. Handles magic and type coercion.
c461cf8f
JH
5970
5971=cut
5972*/
5973
7e8c5dac 5974/*
c05a5c57 5975 * The length is cached in PERL_MAGIC_utf8, in the mg_len field. Also the
9564a3bd
NC
5976 * mg_ptr is used, by sv_pos_u2b() and sv_pos_b2u() - see the comments below.
5977 * (Note that the mg_len is not the length of the mg_ptr field.
5978 * This allows the cache to store the character length of the string without
5979 * needing to malloc() extra storage to attach to the mg_ptr.)
7a5fa8a2 5980 *
7e8c5dac
HS
5981 */
5982
a0ed51b3 5983STRLEN
af828c01 5984Perl_sv_len_utf8(pTHX_ register SV *const sv)
a0ed51b3 5985{
a0ed51b3
LW
5986 if (!sv)
5987 return 0;
5988
a0ed51b3 5989 if (SvGMAGICAL(sv))
b76347f2 5990 return mg_length(sv);
a0ed51b3 5991 else
b76347f2 5992 {
26346457 5993 STRLEN len;
e62f0680 5994 const U8 *s = (U8*)SvPV_const(sv, len);
7e8c5dac 5995
26346457
NC
5996 if (PL_utf8cache) {
5997 STRLEN ulen;
fe5bfecd 5998 MAGIC *mg = SvMAGICAL(sv) ? mg_find(sv, PERL_MAGIC_utf8) : NULL;
26346457
NC
5999
6000 if (mg && mg->mg_len != -1) {
6001 ulen = mg->mg_len;
6002 if (PL_utf8cache < 0) {
6003 const STRLEN real = Perl_utf8_length(aTHX_ s, s + len);
6004 if (real != ulen) {
6005 /* Need to turn the assertions off otherwise we may
6006 recurse infinitely while printing error messages.
6007 */
6008 SAVEI8(PL_utf8cache);
6009 PL_utf8cache = 0;
f5992bc4
RB
6010 Perl_croak(aTHX_ "panic: sv_len_utf8 cache %"UVuf
6011 " real %"UVuf" for %"SVf,
be2597df 6012 (UV) ulen, (UV) real, SVfARG(sv));
26346457
NC
6013 }
6014 }
6015 }
6016 else {
6017 ulen = Perl_utf8_length(aTHX_ s, s + len);
6018 if (!SvREADONLY(sv)) {
f89a570b
CS
6019 if (!mg && (SvTYPE(sv) < SVt_PVMG ||
6020 !(mg = mg_find(sv, PERL_MAGIC_utf8)))) {
26346457
NC
6021 mg = sv_magicext(sv, 0, PERL_MAGIC_utf8,
6022 &PL_vtbl_utf8, 0, 0);
6023 }
cb9e20bb 6024 assert(mg);
26346457 6025 mg->mg_len = ulen;
cb9e20bb 6026 }
cb9e20bb 6027 }
26346457 6028 return ulen;
7e8c5dac 6029 }
26346457 6030 return Perl_utf8_length(aTHX_ s, s + len);
7e8c5dac
HS
6031 }
6032}
6033
9564a3bd
NC
6034/* Walk forwards to find the byte corresponding to the passed in UTF-8
6035 offset. */
bdf30dd6 6036static STRLEN
721e86b6 6037S_sv_pos_u2b_forwards(const U8 *const start, const U8 *const send,
bdf30dd6
NC
6038 STRLEN uoffset)
6039{
6040 const U8 *s = start;
6041
7918f24d
NC
6042 PERL_ARGS_ASSERT_SV_POS_U2B_FORWARDS;
6043
bdf30dd6
NC
6044 while (s < send && uoffset--)
6045 s += UTF8SKIP(s);
6046 if (s > send) {
6047 /* This is the existing behaviour. Possibly it should be a croak, as
6048 it's actually a bounds error */
6049 s = send;
6050 }
6051 return s - start;
6052}
6053
9564a3bd
NC
6054/* Given the length of the string in both bytes and UTF-8 characters, decide
6055 whether to walk forwards or backwards to find the byte corresponding to
6056 the passed in UTF-8 offset. */
c336ad0b 6057static STRLEN
721e86b6 6058S_sv_pos_u2b_midway(const U8 *const start, const U8 *send,
af828c01 6059 const STRLEN uoffset, const STRLEN uend)
c336ad0b
NC
6060{
6061 STRLEN backw = uend - uoffset;
7918f24d
NC
6062
6063 PERL_ARGS_ASSERT_SV_POS_U2B_MIDWAY;
6064
c336ad0b 6065 if (uoffset < 2 * backw) {
25a8a4ef 6066 /* The assumption is that going forwards is twice the speed of going
c336ad0b
NC
6067 forward (that's where the 2 * backw comes from).
6068 (The real figure of course depends on the UTF-8 data.) */
721e86b6 6069 return sv_pos_u2b_forwards(start, send, uoffset);
c336ad0b
NC
6070 }
6071
6072 while (backw--) {
6073 send--;
6074 while (UTF8_IS_CONTINUATION(*send))
6075 send--;
6076 }
6077 return send - start;
6078}
6079
9564a3bd
NC
6080/* For the string representation of the given scalar, find the byte
6081 corresponding to the passed in UTF-8 offset. uoffset0 and boffset0
6082 give another position in the string, *before* the sought offset, which
6083 (which is always true, as 0, 0 is a valid pair of positions), which should
6084 help reduce the amount of linear searching.
6085 If *mgp is non-NULL, it should point to the UTF-8 cache magic, which
6086 will be used to reduce the amount of linear searching. The cache will be
6087 created if necessary, and the found value offered to it for update. */
28ccbf94 6088static STRLEN
af828c01
SS
6089S_sv_pos_u2b_cached(pTHX_ SV *const sv, MAGIC **const mgp, const U8 *const start,
6090 const U8 *const send, const STRLEN uoffset,
7918f24d
NC
6091 STRLEN uoffset0, STRLEN boffset0)
6092{
7087a21c 6093 STRLEN boffset = 0; /* Actually always set, but let's keep gcc happy. */
c336ad0b
NC
6094 bool found = FALSE;
6095
7918f24d
NC
6096 PERL_ARGS_ASSERT_SV_POS_U2B_CACHED;
6097
75c33c12
NC
6098 assert (uoffset >= uoffset0);
6099
f89a570b
CS
6100 if (!SvREADONLY(sv)
6101 && PL_utf8cache
6102 && (*mgp || (SvTYPE(sv) >= SVt_PVMG &&
6103 (*mgp = mg_find(sv, PERL_MAGIC_utf8))))) {
d8b2e1f9
NC
6104 if ((*mgp)->mg_ptr) {
6105 STRLEN *cache = (STRLEN *) (*mgp)->mg_ptr;
6106 if (cache[0] == uoffset) {
6107 /* An exact match. */
6108 return cache[1];
6109 }
ab455f60
NC
6110 if (cache[2] == uoffset) {
6111 /* An exact match. */
6112 return cache[3];
6113 }
668af93f
NC
6114
6115 if (cache[0] < uoffset) {
d8b2e1f9
NC
6116 /* The cache already knows part of the way. */
6117 if (cache[0] > uoffset0) {
6118 /* The cache knows more than the passed in pair */
6119 uoffset0 = cache[0];
6120 boffset0 = cache[1];
6121 }
6122 if ((*mgp)->mg_len != -1) {
6123 /* And we know the end too. */
6124 boffset = boffset0
721e86b6 6125 + sv_pos_u2b_midway(start + boffset0, send,
d8b2e1f9
NC
6126 uoffset - uoffset0,
6127 (*mgp)->mg_len - uoffset0);
6128 } else {
6129 boffset = boffset0
721e86b6 6130 + sv_pos_u2b_forwards(start + boffset0,
d8b2e1f9
NC
6131 send, uoffset - uoffset0);
6132 }
dd7c5fd3
NC
6133 }
6134 else if (cache[2] < uoffset) {
6135 /* We're between the two cache entries. */
6136 if (cache[2] > uoffset0) {
6137 /* and the cache knows more than the passed in pair */
6138 uoffset0 = cache[2];
6139 boffset0 = cache[3];
6140 }
6141
668af93f 6142 boffset = boffset0
721e86b6 6143 + sv_pos_u2b_midway(start + boffset0,
668af93f
NC
6144 start + cache[1],
6145 uoffset - uoffset0,
6146 cache[0] - uoffset0);
dd7c5fd3
NC
6147 } else {
6148 boffset = boffset0
721e86b6 6149 + sv_pos_u2b_midway(start + boffset0,
dd7c5fd3
NC
6150 start + cache[3],
6151 uoffset - uoffset0,
6152 cache[2] - uoffset0);
d8b2e1f9 6153 }
668af93f 6154 found = TRUE;
d8b2e1f9
NC
6155 }
6156 else if ((*mgp)->mg_len != -1) {
75c33c12
NC
6157 /* If we can take advantage of a passed in offset, do so. */
6158 /* In fact, offset0 is either 0, or less than offset, so don't
6159 need to worry about the other possibility. */
6160 boffset = boffset0
721e86b6 6161 + sv_pos_u2b_midway(start + boffset0, send,
75c33c12
NC
6162 uoffset - uoffset0,
6163 (*mgp)->mg_len - uoffset0);
c336ad0b
NC
6164 found = TRUE;
6165 }
28ccbf94 6166 }
c336ad0b
NC
6167
6168 if (!found || PL_utf8cache < 0) {
75c33c12 6169 const STRLEN real_boffset
721e86b6 6170 = boffset0 + sv_pos_u2b_forwards(start + boffset0,
75c33c12
NC
6171 send, uoffset - uoffset0);
6172
c336ad0b
NC
6173 if (found && PL_utf8cache < 0) {
6174 if (real_boffset != boffset) {
6175 /* Need to turn the assertions off otherwise we may recurse
6176 infinitely while printing error messages. */
6177 SAVEI8(PL_utf8cache);
6178 PL_utf8cache = 0;
f5992bc4
RB
6179 Perl_croak(aTHX_ "panic: sv_pos_u2b_cache cache %"UVuf
6180 " real %"UVuf" for %"SVf,
be2597df 6181 (UV) boffset, (UV) real_boffset, SVfARG(sv));
c336ad0b
NC
6182 }
6183 }
6184 boffset = real_boffset;
28ccbf94 6185 }
0905937d 6186
efcbbafb
NC
6187 if (PL_utf8cache)
6188 utf8_mg_pos_cache_update(sv, mgp, boffset, uoffset, send - start);
28ccbf94
NC
6189 return boffset;
6190}
6191
9564a3bd
NC
6192
6193/*
6194=for apidoc sv_pos_u2b
6195
6196Converts the value pointed to by offsetp from a count of UTF-8 chars from
6197the start of the string, to a count of the equivalent number of bytes; if
6198lenp is non-zero, it does the same to lenp, but this time starting from
6199the offset, rather than from the start of the string. Handles magic and
6200type coercion.
6201
6202=cut
6203*/
6204
6205/*
6206 * sv_pos_u2b() uses, like sv_pos_b2u(), the mg_ptr of the potential
c05a5c57 6207 * PERL_MAGIC_utf8 of the sv to store the mapping between UTF-8 and
9564a3bd
NC
6208 * byte offsets. See also the comments of S_utf8_mg_pos_cache_update().
6209 *
6210 */
6211
a0ed51b3 6212void
af828c01 6213Perl_sv_pos_u2b(pTHX_ register SV *const sv, I32 *const offsetp, I32 *const lenp)
a0ed51b3 6214{
245d4a47 6215 const U8 *start;
a0ed51b3
LW
6216 STRLEN len;
6217
7918f24d
NC
6218 PERL_ARGS_ASSERT_SV_POS_U2B;
6219
a0ed51b3
LW
6220 if (!sv)
6221 return;
6222
245d4a47 6223 start = (U8*)SvPV_const(sv, len);
7e8c5dac 6224 if (len) {
bdf30dd6
NC
6225 STRLEN uoffset = (STRLEN) *offsetp;
6226 const U8 * const send = start + len;
0905937d 6227 MAGIC *mg = NULL;
721e86b6 6228 const STRLEN boffset = sv_pos_u2b_cached(sv, &mg, start, send,
28ccbf94 6229 uoffset, 0, 0);
bdf30dd6
NC
6230
6231 *offsetp = (I32) boffset;
6232
6233 if (lenp) {
28ccbf94 6234 /* Convert the relative offset to absolute. */
721e86b6
AL
6235 const STRLEN uoffset2 = uoffset + (STRLEN) *lenp;
6236 const STRLEN boffset2
6237 = sv_pos_u2b_cached(sv, &mg, start, send, uoffset2,
28ccbf94 6238 uoffset, boffset) - boffset;
bdf30dd6 6239
28ccbf94 6240 *lenp = boffset2;
bdf30dd6 6241 }
7e8c5dac
HS
6242 }
6243 else {
6244 *offsetp = 0;
6245 if (lenp)
6246 *lenp = 0;
a0ed51b3 6247 }
e23c8137 6248
a0ed51b3
LW
6249 return;
6250}
6251
9564a3bd
NC
6252/* Create and update the UTF8 magic offset cache, with the proffered utf8/
6253 byte length pairing. The (byte) length of the total SV is passed in too,
6254 as blen, because for some (more esoteric) SVs, the call to SvPV_const()
6255 may not have updated SvCUR, so we can't rely on reading it directly.
6256
6257 The proffered utf8/byte length pairing isn't used if the cache already has
6258 two pairs, and swapping either for the proffered pair would increase the
6259 RMS of the intervals between known byte offsets.
6260
6261 The cache itself consists of 4 STRLEN values
6262 0: larger UTF-8 offset
6263 1: corresponding byte offset
6264 2: smaller UTF-8 offset
6265 3: corresponding byte offset
6266
6267 Unused cache pairs have the value 0, 0.
6268 Keeping the cache "backwards" means that the invariant of
6269 cache[0] >= cache[2] is maintained even with empty slots, which means that
6270 the code that uses it doesn't need to worry if only 1 entry has actually
6271 been set to non-zero. It also makes the "position beyond the end of the
6272 cache" logic much simpler, as the first slot is always the one to start
6273 from.
645c22ef 6274*/
ec07b5e0 6275static void
ac1e9476
SS
6276S_utf8_mg_pos_cache_update(pTHX_ SV *const sv, MAGIC **const mgp, const STRLEN byte,
6277 const STRLEN utf8, const STRLEN blen)
ec07b5e0
NC
6278{
6279 STRLEN *cache;
7918f24d
NC
6280
6281 PERL_ARGS_ASSERT_UTF8_MG_POS_CACHE_UPDATE;
6282
ec07b5e0
NC
6283 if (SvREADONLY(sv))
6284 return;
6285
f89a570b
CS
6286 if (!*mgp && (SvTYPE(sv) < SVt_PVMG ||
6287 !(*mgp = mg_find(sv, PERL_MAGIC_utf8)))) {
ec07b5e0
NC
6288 *mgp = sv_magicext(sv, 0, PERL_MAGIC_utf8, (MGVTBL*)&PL_vtbl_utf8, 0,
6289 0);
6290 (*mgp)->mg_len = -1;
6291 }
6292 assert(*mgp);
6293
6294 if (!(cache = (STRLEN *)(*mgp)->mg_ptr)) {
6295 Newxz(cache, PERL_MAGIC_UTF8_CACHESIZE * 2, STRLEN);
6296 (*mgp)->mg_ptr = (char *) cache;
6297 }
6298 assert(cache);
6299
ab8be49d
NC
6300 if (PL_utf8cache < 0 && SvPOKp(sv)) {
6301 /* SvPOKp() because it's possible that sv has string overloading, and
6302 therefore is a reference, hence SvPVX() is actually a pointer.
6303 This cures the (very real) symptoms of RT 69422, but I'm not actually
6304 sure whether we should even be caching the results of UTF-8
6305 operations on overloading, given that nothing stops overloading
6306 returning a different value every time it's called. */
ef816a78 6307 const U8 *start = (const U8 *) SvPVX_const(sv);
6448472a 6308 const STRLEN realutf8 = utf8_length(start, start + byte);
ec07b5e0
NC
6309
6310 if (realutf8 != utf8) {
6311 /* Need to turn the assertions off otherwise we may recurse
6312 infinitely while printing error messages. */
6313 SAVEI8(PL_utf8cache);
6314 PL_utf8cache = 0;
f5992bc4 6315 Perl_croak(aTHX_ "panic: utf8_mg_pos_cache_update cache %"UVuf
be2597df 6316 " real %"UVuf" for %"SVf, (UV) utf8, (UV) realutf8, SVfARG(sv));
ec07b5e0
NC
6317 }
6318 }
ab455f60
NC
6319
6320 /* Cache is held with the later position first, to simplify the code
6321 that deals with unbounded ends. */
6322
6323 ASSERT_UTF8_CACHE(cache);
6324 if (cache[1] == 0) {
6325 /* Cache is totally empty */
6326 cache[0] = utf8;
6327 cache[1] = byte;
6328 } else if (cache[3] == 0) {
6329 if (byte > cache[1]) {
6330 /* New one is larger, so goes first. */
6331 cache[2] = cache[0];
6332 cache[3] = cache[1];
6333 cache[0] = utf8;
6334 cache[1] = byte;
6335 } else {
6336 cache[2] = utf8;
6337 cache[3] = byte;
6338 }
6339 } else {
6340#define THREEWAY_SQUARE(a,b,c,d) \
6341 ((float)((d) - (c))) * ((float)((d) - (c))) \
6342 + ((float)((c) - (b))) * ((float)((c) - (b))) \
6343 + ((float)((b) - (a))) * ((float)((b) - (a)))
6344
6345 /* Cache has 2 slots in use, and we know three potential pairs.
6346 Keep the two that give the lowest RMS distance. Do the
6347 calcualation in bytes simply because we always know the byte
6348 length. squareroot has the same ordering as the positive value,
6349 so don't bother with the actual square root. */
6350 const float existing = THREEWAY_SQUARE(0, cache[3], cache[1], blen);
6351 if (byte > cache[1]) {
6352 /* New position is after the existing pair of pairs. */
6353 const float keep_earlier
6354 = THREEWAY_SQUARE(0, cache[3], byte, blen);
6355 const float keep_later
6356 = THREEWAY_SQUARE(0, cache[1], byte, blen);
6357
6358 if (keep_later < keep_earlier) {
6359 if (keep_later < existing) {
6360 cache[2] = cache[0];
6361 cache[3] = cache[1];
6362 cache[0] = utf8;
6363 cache[1] = byte;
6364 }
6365 }
6366 else {
6367 if (keep_earlier < existing) {
6368 cache[0] = utf8;
6369 cache[1] = byte;
6370 }
6371 }
6372 }
57d7fbf1
NC
6373 else if (byte > cache[3]) {
6374 /* New position is between the existing pair of pairs. */
6375 const float keep_earlier
6376 = THREEWAY_SQUARE(0, cache[3], byte, blen);
6377 const float keep_later
6378 = THREEWAY_SQUARE(0, byte, cache[1], blen);
6379
6380 if (keep_later < keep_earlier) {
6381 if (keep_later < existing) {
6382 cache[2] = utf8;
6383 cache[3] = byte;
6384 }
6385 }
6386 else {
6387 if (keep_earlier < existing) {
6388 cache[0] = utf8;
6389 cache[1] = byte;
6390 }
6391 }
6392 }
6393 else {
6394 /* New position is before the existing pair of pairs. */
6395 const float keep_earlier
6396 = THREEWAY_SQUARE(0, byte, cache[3], blen);
6397 const float keep_later
6398 = THREEWAY_SQUARE(0, byte, cache[1], blen);
6399
6400 if (keep_later < keep_earlier) {
6401 if (keep_later < existing) {
6402 cache[2] = utf8;
6403 cache[3] = byte;
6404 }
6405 }
6406 else {
6407 if (keep_earlier < existing) {
6408 cache[0] = cache[2];
6409 cache[1] = cache[3];
6410 cache[2] = utf8;
6411 cache[3] = byte;
6412 }
6413 }
6414 }
ab455f60 6415 }
0905937d 6416 ASSERT_UTF8_CACHE(cache);
ec07b5e0
NC
6417}
6418
ec07b5e0 6419/* We already know all of the way, now we may be able to walk back. The same
25a8a4ef
NC
6420 assumption is made as in S_sv_pos_u2b_midway(), namely that walking
6421 backward is half the speed of walking forward. */
ec07b5e0 6422static STRLEN
ac1e9476
SS
6423S_sv_pos_b2u_midway(pTHX_ const U8 *const s, const U8 *const target,
6424 const U8 *end, STRLEN endu)
ec07b5e0
NC
6425{
6426 const STRLEN forw = target - s;
6427 STRLEN backw = end - target;
6428
7918f24d
NC
6429 PERL_ARGS_ASSERT_SV_POS_B2U_MIDWAY;
6430
ec07b5e0 6431 if (forw < 2 * backw) {
6448472a 6432 return utf8_length(s, target);
ec07b5e0
NC
6433 }
6434
6435 while (end > target) {
6436 end--;
6437 while (UTF8_IS_CONTINUATION(*end)) {
6438 end--;
6439 }
6440 endu--;
6441 }
6442 return endu;
6443}
6444
9564a3bd
NC
6445/*
6446=for apidoc sv_pos_b2u
6447
6448Converts the value pointed to by offsetp from a count of bytes from the
6449start of the string, to a count of the equivalent number of UTF-8 chars.
6450Handles magic and type coercion.
6451
6452=cut
6453*/
6454
6455/*
6456 * sv_pos_b2u() uses, like sv_pos_u2b(), the mg_ptr of the potential
c05a5c57 6457 * PERL_MAGIC_utf8 of the sv to store the mapping between UTF-8 and
9564a3bd
NC
6458 * byte offsets.
6459 *
6460 */
a0ed51b3 6461void
ac1e9476 6462Perl_sv_pos_b2u(pTHX_ register SV *const sv, I32 *const offsetp)
a0ed51b3 6463{
83003860 6464 const U8* s;
ec07b5e0 6465 const STRLEN byte = *offsetp;
7087a21c 6466 STRLEN len = 0; /* Actually always set, but let's keep gcc happy. */
ab455f60 6467 STRLEN blen;
ec07b5e0
NC
6468 MAGIC* mg = NULL;
6469 const U8* send;
a922f900 6470 bool found = FALSE;
a0ed51b3 6471
7918f24d
NC
6472 PERL_ARGS_ASSERT_SV_POS_B2U;
6473
a0ed51b3
LW
6474 if (!sv)
6475 return;
6476
ab455f60 6477 s = (const U8*)SvPV_const(sv, blen);
7e8c5dac 6478
ab455f60 6479 if (blen < byte)
ec07b5e0 6480 Perl_croak(aTHX_ "panic: sv_pos_b2u: bad byte offset");
7e8c5dac 6481
ec07b5e0 6482 send = s + byte;
a67d7df9 6483
f89a570b
CS
6484 if (!SvREADONLY(sv)
6485 && PL_utf8cache
6486 && SvTYPE(sv) >= SVt_PVMG
6487 && (mg = mg_find(sv, PERL_MAGIC_utf8)))
6488 {
ffca234a 6489 if (mg->mg_ptr) {
d4c19fe8 6490 STRLEN * const cache = (STRLEN *) mg->mg_ptr;
b9f984a5 6491 if (cache[1] == byte) {
ec07b5e0
NC
6492 /* An exact match. */
6493 *offsetp = cache[0];
ec07b5e0 6494 return;
7e8c5dac 6495 }
ab455f60
NC
6496 if (cache[3] == byte) {
6497 /* An exact match. */
6498 *offsetp = cache[2];
6499 return;
6500 }
668af93f
NC
6501
6502 if (cache[1] < byte) {
ec07b5e0 6503 /* We already know part of the way. */
b9f984a5
NC
6504 if (mg->mg_len != -1) {
6505 /* Actually, we know the end too. */
6506 len = cache[0]
6507 + S_sv_pos_b2u_midway(aTHX_ s + cache[1], send,
ab455f60 6508 s + blen, mg->mg_len - cache[0]);
b9f984a5 6509 } else {
6448472a 6510 len = cache[0] + utf8_length(s + cache[1], send);
b9f984a5 6511 }
7e8c5dac 6512 }
9f985e4c
NC
6513 else if (cache[3] < byte) {
6514 /* We're between the two cached pairs, so we do the calculation
6515 offset by the byte/utf-8 positions for the earlier pair,
6516 then add the utf-8 characters from the string start to
6517 there. */
6518 len = S_sv_pos_b2u_midway(aTHX_ s + cache[3], send,
6519 s + cache[1], cache[0] - cache[2])
6520 + cache[2];
6521
6522 }
6523 else { /* cache[3] > byte */
6524 len = S_sv_pos_b2u_midway(aTHX_ s, send, s + cache[3],
6525 cache[2]);
7e8c5dac 6526
7e8c5dac 6527 }
ec07b5e0 6528 ASSERT_UTF8_CACHE(cache);
a922f900 6529 found = TRUE;
ffca234a 6530 } else if (mg->mg_len != -1) {
ab455f60 6531 len = S_sv_pos_b2u_midway(aTHX_ s, send, s + blen, mg->mg_len);
a922f900 6532 found = TRUE;
7e8c5dac 6533 }
a0ed51b3 6534 }
a922f900 6535 if (!found || PL_utf8cache < 0) {
6448472a 6536 const STRLEN real_len = utf8_length(s, send);
a922f900
NC
6537
6538 if (found && PL_utf8cache < 0) {
6539 if (len != real_len) {
6540 /* Need to turn the assertions off otherwise we may recurse
6541 infinitely while printing error messages. */
6542 SAVEI8(PL_utf8cache);
6543 PL_utf8cache = 0;
f5992bc4
RB
6544 Perl_croak(aTHX_ "panic: sv_pos_b2u cache %"UVuf
6545 " real %"UVuf" for %"SVf,
be2597df 6546 (UV) len, (UV) real_len, SVfARG(sv));
a922f900
NC
6547 }
6548 }
6549 len = real_len;
ec07b5e0
NC
6550 }
6551 *offsetp = len;
6552
efcbbafb
NC
6553 if (PL_utf8cache)
6554 utf8_mg_pos_cache_update(sv, &mg, byte, len, blen);
a0ed51b3
LW
6555}
6556
954c1994
GS
6557/*
6558=for apidoc sv_eq
6559
6560Returns a boolean indicating whether the strings in the two SVs are
645c22ef
DM
6561identical. Is UTF-8 and 'use bytes' aware, handles get magic, and will
6562coerce its args to strings if necessary.
954c1994
GS
6563
6564=cut
6565*/
6566
79072805 6567I32
e01b9e88 6568Perl_sv_eq(pTHX_ register SV *sv1, register SV *sv2)
79072805 6569{
97aff369 6570 dVAR;
e1ec3a88 6571 const char *pv1;
463ee0b2 6572 STRLEN cur1;
e1ec3a88 6573 const char *pv2;
463ee0b2 6574 STRLEN cur2;
e01b9e88 6575 I32 eq = 0;
bd61b366 6576 char *tpv = NULL;
a0714e2c 6577 SV* svrecode = NULL;
79072805 6578
e01b9e88 6579 if (!sv1) {
79072805
LW
6580 pv1 = "";
6581 cur1 = 0;
6582 }
ced497e2
YST
6583 else {
6584 /* if pv1 and pv2 are the same, second SvPV_const call may
6585 * invalidate pv1, so we may need to make a copy */
6586 if (sv1 == sv2 && (SvTHINKFIRST(sv1) || SvGMAGICAL(sv1))) {
6587 pv1 = SvPV_const(sv1, cur1);
59cd0e26 6588 sv1 = newSVpvn_flags(pv1, cur1, SVs_TEMP | SvUTF8(sv2));
ced497e2 6589 }
4d84ee25 6590 pv1 = SvPV_const(sv1, cur1);
ced497e2 6591 }
79072805 6592
e01b9e88
SC
6593 if (!sv2){
6594 pv2 = "";
6595 cur2 = 0;
92d29cee 6596 }
e01b9e88 6597 else
4d84ee25 6598 pv2 = SvPV_const(sv2, cur2);
79072805 6599
cf48d248 6600 if (cur1 && cur2 && SvUTF8(sv1) != SvUTF8(sv2) && !IN_BYTES) {
799ef3cb
JH
6601 /* Differing utf8ness.
6602 * Do not UTF8size the comparands as a side-effect. */
6603 if (PL_encoding) {
6604 if (SvUTF8(sv1)) {
553e1bcc
AT
6605 svrecode = newSVpvn(pv2, cur2);
6606 sv_recode_to_utf8(svrecode, PL_encoding);
93524f2b 6607 pv2 = SvPV_const(svrecode, cur2);
799ef3cb
JH
6608 }
6609 else {
553e1bcc
AT
6610 svrecode = newSVpvn(pv1, cur1);
6611 sv_recode_to_utf8(svrecode, PL_encoding);
93524f2b 6612 pv1 = SvPV_const(svrecode, cur1);
799ef3cb
JH
6613 }
6614 /* Now both are in UTF-8. */
0a1bd7ac
DM
6615 if (cur1 != cur2) {
6616 SvREFCNT_dec(svrecode);
799ef3cb 6617 return FALSE;
0a1bd7ac 6618 }
799ef3cb
JH
6619 }
6620 else {
6621 bool is_utf8 = TRUE;
6622
6623 if (SvUTF8(sv1)) {
6624 /* sv1 is the UTF-8 one,
6625 * if is equal it must be downgrade-able */
9d4ba2ae 6626 char * const pv = (char*)bytes_from_utf8((const U8*)pv1,
799ef3cb
JH
6627 &cur1, &is_utf8);
6628 if (pv != pv1)
553e1bcc 6629 pv1 = tpv = pv;
799ef3cb
JH
6630 }
6631 else {
6632 /* sv2 is the UTF-8 one,
6633 * if is equal it must be downgrade-able */
9d4ba2ae 6634 char * const pv = (char *)bytes_from_utf8((const U8*)pv2,
799ef3cb
JH
6635 &cur2, &is_utf8);
6636 if (pv != pv2)
553e1bcc 6637 pv2 = tpv = pv;
799ef3cb
JH
6638 }
6639 if (is_utf8) {
6640 /* Downgrade not possible - cannot be eq */
bf694877 6641 assert (tpv == 0);
799ef3cb
JH
6642 return FALSE;
6643 }
6644 }
cf48d248
JH
6645 }
6646
6647 if (cur1 == cur2)
765f542d 6648 eq = (pv1 == pv2) || memEQ(pv1, pv2, cur1);
e01b9e88 6649
b37c2d43 6650 SvREFCNT_dec(svrecode);
553e1bcc
AT
6651 if (tpv)
6652 Safefree(tpv);
cf48d248 6653
e01b9e88 6654 return eq;
79072805
LW
6655}
6656
954c1994
GS
6657/*
6658=for apidoc sv_cmp
6659
6660Compares the strings in two SVs. Returns -1, 0, or 1 indicating whether the
6661string in C<sv1> is less than, equal to, or greater than the string in
645c22ef
DM
6662C<sv2>. Is UTF-8 and 'use bytes' aware, handles get magic, and will
6663coerce its args to strings if necessary. See also C<sv_cmp_locale>.
954c1994
GS
6664
6665=cut
6666*/
6667
79072805 6668I32
ac1e9476 6669Perl_sv_cmp(pTHX_ register SV *const sv1, register SV *const sv2)
79072805 6670{
97aff369 6671 dVAR;
560a288e 6672 STRLEN cur1, cur2;
e1ec3a88 6673 const char *pv1, *pv2;
bd61b366 6674 char *tpv = NULL;
cf48d248 6675 I32 cmp;
a0714e2c 6676 SV *svrecode = NULL;
560a288e 6677
e01b9e88
SC
6678 if (!sv1) {
6679 pv1 = "";
560a288e
GS
6680 cur1 = 0;
6681 }
e01b9e88 6682 else
4d84ee25 6683 pv1 = SvPV_const(sv1, cur1);
560a288e 6684
553e1bcc 6685 if (!sv2) {
e01b9e88 6686 pv2 = "";
560a288e
GS
6687 cur2 = 0;
6688 }
e01b9e88 6689 else
4d84ee25 6690 pv2 = SvPV_const(sv2, cur2);
79072805 6691
cf48d248 6692 if (cur1 && cur2 && SvUTF8(sv1) != SvUTF8(sv2) && !IN_BYTES) {
799ef3cb
JH
6693 /* Differing utf8ness.
6694 * Do not UTF8size the comparands as a side-effect. */
cf48d248 6695 if (SvUTF8(sv1)) {
799ef3cb 6696 if (PL_encoding) {
553e1bcc
AT
6697 svrecode = newSVpvn(pv2, cur2);
6698 sv_recode_to_utf8(svrecode, PL_encoding);
93524f2b 6699 pv2 = SvPV_const(svrecode, cur2);
799ef3cb
JH
6700 }
6701 else {
e1ec3a88 6702 pv2 = tpv = (char*)bytes_to_utf8((const U8*)pv2, &cur2);
799ef3cb 6703 }
cf48d248
JH
6704 }
6705 else {
799ef3cb 6706 if (PL_encoding) {
553e1bcc
AT
6707 svrecode = newSVpvn(pv1, cur1);
6708 sv_recode_to_utf8(svrecode, PL_encoding);
93524f2b 6709 pv1 = SvPV_const(svrecode, cur1);
799ef3cb
JH
6710 }
6711 else {
e1ec3a88 6712 pv1 = tpv = (char*)bytes_to_utf8((const U8*)pv1, &cur1);
799ef3cb 6713 }
cf48d248
JH
6714 }
6715 }
6716
e01b9e88 6717 if (!cur1) {
cf48d248 6718 cmp = cur2 ? -1 : 0;
e01b9e88 6719 } else if (!cur2) {
cf48d248
JH
6720 cmp = 1;
6721 } else {
e1ec3a88 6722 const I32 retval = memcmp((const void*)pv1, (const void*)pv2, cur1 < cur2 ? cur1 : cur2);
e01b9e88
SC
6723
6724 if (retval) {
cf48d248 6725 cmp = retval < 0 ? -1 : 1;
e01b9e88 6726 } else if (cur1 == cur2) {
cf48d248
JH
6727 cmp = 0;
6728 } else {
6729 cmp = cur1 < cur2 ? -1 : 1;
e01b9e88 6730 }
cf48d248 6731 }
16660edb 6732
b37c2d43 6733 SvREFCNT_dec(svrecode);
553e1bcc
AT
6734 if (tpv)
6735 Safefree(tpv);
cf48d248
JH
6736
6737 return cmp;
bbce6d69 6738}
16660edb 6739
c461cf8f
JH
6740/*
6741=for apidoc sv_cmp_locale
6742
645c22ef
DM
6743Compares the strings in two SVs in a locale-aware manner. Is UTF-8 and
6744'use bytes' aware, handles get magic, and will coerce its args to strings
d77cdebf 6745if necessary. See also C<sv_cmp>.
c461cf8f
JH
6746
6747=cut
6748*/
6749
bbce6d69 6750I32
ac1e9476 6751Perl_sv_cmp_locale(pTHX_ register SV *const sv1, register SV *const sv2)
bbce6d69 6752{
97aff369 6753 dVAR;
36477c24 6754#ifdef USE_LOCALE_COLLATE
16660edb 6755
bbce6d69 6756 char *pv1, *pv2;
6757 STRLEN len1, len2;
6758 I32 retval;
16660edb 6759
3280af22 6760 if (PL_collation_standard)
bbce6d69 6761 goto raw_compare;
16660edb 6762
bbce6d69 6763 len1 = 0;
8ac85365 6764 pv1 = sv1 ? sv_collxfrm(sv1, &len1) : (char *) NULL;
bbce6d69 6765 len2 = 0;
8ac85365 6766 pv2 = sv2 ? sv_collxfrm(sv2, &len2) : (char *) NULL;
16660edb 6767
bbce6d69 6768 if (!pv1 || !len1) {
6769 if (pv2 && len2)
6770 return -1;
6771 else
6772 goto raw_compare;
6773 }
6774 else {
6775 if (!pv2 || !len2)
6776 return 1;
6777 }
16660edb 6778
bbce6d69 6779 retval = memcmp((void*)pv1, (void*)pv2, len1 < len2 ? len1 : len2);
16660edb 6780
bbce6d69 6781 if (retval)
16660edb 6782 return retval < 0 ? -1 : 1;
6783
bbce6d69 6784 /*
6785 * When the result of collation is equality, that doesn't mean
6786 * that there are no differences -- some locales exclude some
6787 * characters from consideration. So to avoid false equalities,
6788 * we use the raw string as a tiebreaker.
6789 */
16660edb 6790
bbce6d69 6791 raw_compare:
5f66b61c 6792 /*FALLTHROUGH*/
16660edb 6793
36477c24 6794#endif /* USE_LOCALE_COLLATE */
16660edb 6795
bbce6d69 6796 return sv_cmp(sv1, sv2);
6797}
79072805 6798
645c22ef 6799
36477c24 6800#ifdef USE_LOCALE_COLLATE
645c22ef 6801
7a4c00b4 6802/*
645c22ef
DM
6803=for apidoc sv_collxfrm
6804
6805Add Collate Transform magic to an SV if it doesn't already have it.
6806
6807Any scalar variable may carry PERL_MAGIC_collxfrm magic that contains the
6808scalar data of the variable, but transformed to such a format that a normal
6809memory comparison can be used to compare the data according to the locale
6810settings.
6811
6812=cut
6813*/
6814
bbce6d69 6815char *
ac1e9476 6816Perl_sv_collxfrm(pTHX_ SV *const sv, STRLEN *const nxp)
bbce6d69 6817{
97aff369 6818 dVAR;
7a4c00b4 6819 MAGIC *mg;
16660edb 6820
7918f24d
NC
6821 PERL_ARGS_ASSERT_SV_COLLXFRM;
6822
14befaf4 6823 mg = SvMAGICAL(sv) ? mg_find(sv, PERL_MAGIC_collxfrm) : (MAGIC *) NULL;
3280af22 6824 if (!mg || !mg->mg_ptr || *(U32*)mg->mg_ptr != PL_collation_ix) {
93524f2b
NC
6825 const char *s;
6826 char *xf;
bbce6d69 6827 STRLEN len, xlen;
6828
7a4c00b4 6829 if (mg)
6830 Safefree(mg->mg_ptr);
93524f2b 6831 s = SvPV_const(sv, len);
bbce6d69 6832 if ((xf = mem_collxfrm(s, len, &xlen))) {
7a4c00b4 6833 if (! mg) {
d83f0a82
NC
6834#ifdef PERL_OLD_COPY_ON_WRITE
6835 if (SvIsCOW(sv))
6836 sv_force_normal_flags(sv, 0);
6837#endif
6838 mg = sv_magicext(sv, 0, PERL_MAGIC_collxfrm, &PL_vtbl_collxfrm,
6839 0, 0);
7a4c00b4 6840 assert(mg);
bbce6d69 6841 }
7a4c00b4 6842 mg->mg_ptr = xf;
565764a8 6843 mg->mg_len = xlen;
7a4c00b4 6844 }
6845 else {
ff0cee69 6846 if (mg) {
6847 mg->mg_ptr = NULL;
565764a8 6848 mg->mg_len = -1;
ff0cee69 6849 }
bbce6d69 6850 }
6851 }
7a4c00b4 6852 if (mg && mg->mg_ptr) {
565764a8 6853 *nxp = mg->mg_len;
3280af22 6854 return mg->mg_ptr + sizeof(PL_collation_ix);
bbce6d69 6855 }
6856 else {
6857 *nxp = 0;
6858 return NULL;
16660edb 6859 }
79072805
LW
6860}
6861
36477c24 6862#endif /* USE_LOCALE_COLLATE */
bbce6d69 6863
c461cf8f
JH
6864/*
6865=for apidoc sv_gets
6866
6867Get a line from the filehandle and store it into the SV, optionally
6868appending to the currently-stored string.
6869
6870=cut
6871*/
6872
79072805 6873char *
ac1e9476 6874Perl_sv_gets(pTHX_ register SV *const sv, register PerlIO *const fp, I32 append)
79072805 6875{
97aff369 6876 dVAR;
e1ec3a88 6877 const char *rsptr;
c07a80fd 6878 STRLEN rslen;
6879 register STDCHAR rslast;
6880 register STDCHAR *bp;
6881 register I32 cnt;
9c5ffd7c 6882 I32 i = 0;
8bfdd7d9 6883 I32 rspara = 0;
c07a80fd 6884
7918f24d
NC
6885 PERL_ARGS_ASSERT_SV_GETS;
6886
bc44a8a2
NC
6887 if (SvTHINKFIRST(sv))
6888 sv_force_normal_flags(sv, append ? 0 : SV_COW_DROP_PV);
765f542d
NC
6889 /* XXX. If you make this PVIV, then copy on write can copy scalars read
6890 from <>.
6891 However, perlbench says it's slower, because the existing swipe code
6892 is faster than copy on write.
6893 Swings and roundabouts. */
862a34c6 6894 SvUPGRADE(sv, SVt_PV);
99491443 6895
ff68c719 6896 SvSCREAM_off(sv);
efd8b2ba
AE
6897
6898 if (append) {
6899 if (PerlIO_isutf8(fp)) {
6900 if (!SvUTF8(sv)) {
6901 sv_utf8_upgrade_nomg(sv);
6902 sv_pos_u2b(sv,&append,0);
6903 }
6904 } else if (SvUTF8(sv)) {
561b68a9 6905 SV * const tsv = newSV(0);
efd8b2ba
AE
6906 sv_gets(tsv, fp, 0);
6907 sv_utf8_upgrade_nomg(tsv);
6908 SvCUR_set(sv,append);
6909 sv_catsv(sv,tsv);
6910 sv_free(tsv);
6911 goto return_string_or_null;
6912 }
6913 }
6914
6915 SvPOK_only(sv);
6916 if (PerlIO_isutf8(fp))
6917 SvUTF8_on(sv);
c07a80fd 6918
923e4eb5 6919 if (IN_PERL_COMPILETIME) {
8bfdd7d9
HS
6920 /* we always read code in line mode */
6921 rsptr = "\n";
6922 rslen = 1;
6923 }
6924 else if (RsSNARF(PL_rs)) {
7a5fa8a2 6925 /* If it is a regular disk file use size from stat() as estimate
acbd132f
JH
6926 of amount we are going to read -- may result in mallocing
6927 more memory than we really need if the layers below reduce
6928 the size we read (e.g. CRLF or a gzip layer).
e468d35b 6929 */
e311fd51 6930 Stat_t st;
e468d35b 6931 if (!PerlLIO_fstat(PerlIO_fileno(fp), &st) && S_ISREG(st.st_mode)) {
f54cb97a 6932 const Off_t offset = PerlIO_tell(fp);
58f1856e 6933 if (offset != (Off_t) -1 && st.st_size + append > offset) {
e468d35b
NIS
6934 (void) SvGROW(sv, (STRLEN)((st.st_size - offset) + append + 1));
6935 }
6936 }
c07a80fd 6937 rsptr = NULL;
6938 rslen = 0;
6939 }
3280af22 6940 else if (RsRECORD(PL_rs)) {
e311fd51 6941 I32 bytesread;
5b2b9c68 6942 char *buffer;
acbd132f 6943 U32 recsize;
048d9da8
CB
6944#ifdef VMS
6945 int fd;
6946#endif
5b2b9c68
HM
6947
6948 /* Grab the size of the record we're getting */
acbd132f 6949 recsize = SvUV(SvRV(PL_rs)); /* RsRECORD() guarantees > 0. */
e311fd51 6950 buffer = SvGROW(sv, (STRLEN)(recsize + append + 1)) + append;
5b2b9c68
HM
6951 /* Go yank in */
6952#ifdef VMS
6953 /* VMS wants read instead of fread, because fread doesn't respect */
6954 /* RMS record boundaries. This is not necessarily a good thing to be */
e468d35b
NIS
6955 /* doing, but we've got no other real choice - except avoid stdio
6956 as implementation - perhaps write a :vms layer ?
6957 */
048d9da8
CB
6958 fd = PerlIO_fileno(fp);
6959 if (fd == -1) { /* in-memory file from PerlIO::Scalar */
6960 bytesread = PerlIO_read(fp, buffer, recsize);
6961 }
6962 else {
6963 bytesread = PerlLIO_read(fd, buffer, recsize);
6964 }
5b2b9c68
HM
6965#else
6966 bytesread = PerlIO_read(fp, buffer, recsize);
6967#endif
27e6ca2d
AE
6968 if (bytesread < 0)
6969 bytesread = 0;
82f1394b 6970 SvCUR_set(sv, bytesread + append);
e670df4e 6971 buffer[bytesread] = '\0';
efd8b2ba 6972 goto return_string_or_null;
5b2b9c68 6973 }
3280af22 6974 else if (RsPARA(PL_rs)) {
c07a80fd 6975 rsptr = "\n\n";
6976 rslen = 2;
8bfdd7d9 6977 rspara = 1;
c07a80fd 6978 }
7d59b7e4
NIS
6979 else {
6980 /* Get $/ i.e. PL_rs into same encoding as stream wants */
6981 if (PerlIO_isutf8(fp)) {
6982 rsptr = SvPVutf8(PL_rs, rslen);
6983 }
6984 else {
6985 if (SvUTF8(PL_rs)) {
6986 if (!sv_utf8_downgrade(PL_rs, TRUE)) {
6987 Perl_croak(aTHX_ "Wide character in $/");
6988 }
6989 }
93524f2b 6990 rsptr = SvPV_const(PL_rs, rslen);
7d59b7e4
NIS
6991 }
6992 }
6993
c07a80fd 6994 rslast = rslen ? rsptr[rslen - 1] : '\0';
6995
8bfdd7d9 6996 if (rspara) { /* have to do this both before and after */
79072805 6997 do { /* to make sure file boundaries work right */
760ac839 6998 if (PerlIO_eof(fp))
a0d0e21e 6999 return 0;
760ac839 7000 i = PerlIO_getc(fp);
79072805 7001 if (i != '\n') {
a0d0e21e
LW
7002 if (i == -1)
7003 return 0;
760ac839 7004 PerlIO_ungetc(fp,i);
79072805
LW
7005 break;
7006 }
7007 } while (i != EOF);
7008 }
c07a80fd 7009
760ac839
LW
7010 /* See if we know enough about I/O mechanism to cheat it ! */
7011
7012 /* This used to be #ifdef test - it is made run-time test for ease
1c846c1f 7013 of abstracting out stdio interface. One call should be cheap
760ac839
LW
7014 enough here - and may even be a macro allowing compile
7015 time optimization.
7016 */
7017
7018 if (PerlIO_fast_gets(fp)) {
7019
7020 /*
7021 * We're going to steal some values from the stdio struct
7022 * and put EVERYTHING in the innermost loop into registers.
7023 */
7024 register STDCHAR *ptr;
7025 STRLEN bpx;
7026 I32 shortbuffered;
7027
16660edb 7028#if defined(VMS) && defined(PERLIO_IS_STDIO)
7029 /* An ungetc()d char is handled separately from the regular
7030 * buffer, so we getc() it back out and stuff it in the buffer.
7031 */
7032 i = PerlIO_getc(fp);
7033 if (i == EOF) return 0;
7034 *(--((*fp)->_ptr)) = (unsigned char) i;
7035 (*fp)->_cnt++;
7036#endif
c07a80fd 7037
c2960299 7038 /* Here is some breathtakingly efficient cheating */
c07a80fd 7039
a20bf0c3 7040 cnt = PerlIO_get_cnt(fp); /* get count into register */
e468d35b 7041 /* make sure we have the room */
7a5fa8a2 7042 if ((I32)(SvLEN(sv) - append) <= cnt + 1) {
e468d35b 7043 /* Not room for all of it
7a5fa8a2 7044 if we are looking for a separator and room for some
e468d35b
NIS
7045 */
7046 if (rslen && cnt > 80 && (I32)SvLEN(sv) > append) {
7a5fa8a2 7047 /* just process what we have room for */
79072805
LW
7048 shortbuffered = cnt - SvLEN(sv) + append + 1;
7049 cnt -= shortbuffered;
7050 }
7051 else {
7052 shortbuffered = 0;
bbce6d69 7053 /* remember that cnt can be negative */
eb160463 7054 SvGROW(sv, (STRLEN)(append + (cnt <= 0 ? 2 : (cnt + 1))));
79072805
LW
7055 }
7056 }
7a5fa8a2 7057 else
79072805 7058 shortbuffered = 0;
3f7c398e 7059 bp = (STDCHAR*)SvPVX_const(sv) + append; /* move these two too to registers */
a20bf0c3 7060 ptr = (STDCHAR*)PerlIO_get_ptr(fp);
16660edb 7061 DEBUG_P(PerlIO_printf(Perl_debug_log,
1d7c1841 7062 "Screamer: entering, ptr=%"UVuf", cnt=%ld\n",PTR2UV(ptr),(long)cnt));
16660edb 7063 DEBUG_P(PerlIO_printf(Perl_debug_log,
ba7abf9d 7064 "Screamer: entering: PerlIO * thinks ptr=%"UVuf", cnt=%ld, base=%"UVuf"\n",
1c846c1f 7065 PTR2UV(PerlIO_get_ptr(fp)), (long)PerlIO_get_cnt(fp),
1d7c1841 7066 PTR2UV(PerlIO_has_base(fp) ? PerlIO_get_base(fp) : 0)));
79072805
LW
7067 for (;;) {
7068 screamer:
93a17b20 7069 if (cnt > 0) {
c07a80fd 7070 if (rslen) {
760ac839
LW
7071 while (cnt > 0) { /* this | eat */
7072 cnt--;
c07a80fd 7073 if ((*bp++ = *ptr++) == rslast) /* really | dust */
7074 goto thats_all_folks; /* screams | sed :-) */
7075 }
7076 }
7077 else {
1c846c1f
NIS
7078 Copy(ptr, bp, cnt, char); /* this | eat */
7079 bp += cnt; /* screams | dust */
c07a80fd 7080 ptr += cnt; /* louder | sed :-) */
a5f75d66 7081 cnt = 0;
93a17b20 7082 }
79072805
LW
7083 }
7084
748a9306 7085 if (shortbuffered) { /* oh well, must extend */
79072805
LW
7086 cnt = shortbuffered;
7087 shortbuffered = 0;
3f7c398e 7088 bpx = bp - (STDCHAR*)SvPVX_const(sv); /* box up before relocation */
79072805
LW
7089 SvCUR_set(sv, bpx);
7090 SvGROW(sv, SvLEN(sv) + append + cnt + 2);
3f7c398e 7091 bp = (STDCHAR*)SvPVX_const(sv) + bpx; /* unbox after relocation */
79072805
LW
7092 continue;
7093 }
7094
16660edb 7095 DEBUG_P(PerlIO_printf(Perl_debug_log,
1d7c1841
GS
7096 "Screamer: going to getc, ptr=%"UVuf", cnt=%ld\n",
7097 PTR2UV(ptr),(long)cnt));
cc00df79 7098 PerlIO_set_ptrcnt(fp, (STDCHAR*)ptr, cnt); /* deregisterize cnt and ptr */
ba7abf9d 7099#if 0
16660edb 7100 DEBUG_P(PerlIO_printf(Perl_debug_log,
1d7c1841 7101 "Screamer: pre: FILE * thinks ptr=%"UVuf", cnt=%ld, base=%"UVuf"\n",
1c846c1f 7102 PTR2UV(PerlIO_get_ptr(fp)), (long)PerlIO_get_cnt(fp),
1d7c1841 7103 PTR2UV(PerlIO_has_base (fp) ? PerlIO_get_base(fp) : 0)));
ba7abf9d 7104#endif
1c846c1f 7105 /* This used to call 'filbuf' in stdio form, but as that behaves like
774d564b 7106 getc when cnt <= 0 we use PerlIO_getc here to avoid introducing
7107 another abstraction. */
760ac839 7108 i = PerlIO_getc(fp); /* get more characters */
ba7abf9d 7109#if 0
16660edb 7110 DEBUG_P(PerlIO_printf(Perl_debug_log,
1d7c1841 7111 "Screamer: post: FILE * thinks ptr=%"UVuf", cnt=%ld, base=%"UVuf"\n",
1c846c1f 7112 PTR2UV(PerlIO_get_ptr(fp)), (long)PerlIO_get_cnt(fp),
1d7c1841 7113 PTR2UV(PerlIO_has_base (fp) ? PerlIO_get_base(fp) : 0)));
ba7abf9d 7114#endif
a20bf0c3
JH
7115 cnt = PerlIO_get_cnt(fp);
7116 ptr = (STDCHAR*)PerlIO_get_ptr(fp); /* reregisterize cnt and ptr */
16660edb 7117 DEBUG_P(PerlIO_printf(Perl_debug_log,
1d7c1841 7118 "Screamer: after getc, ptr=%"UVuf", cnt=%ld\n",PTR2UV(ptr),(long)cnt));
79072805 7119
748a9306
LW
7120 if (i == EOF) /* all done for ever? */
7121 goto thats_really_all_folks;
7122
3f7c398e 7123 bpx = bp - (STDCHAR*)SvPVX_const(sv); /* box up before relocation */
79072805
LW
7124 SvCUR_set(sv, bpx);
7125 SvGROW(sv, bpx + cnt + 2);
3f7c398e 7126 bp = (STDCHAR*)SvPVX_const(sv) + bpx; /* unbox after relocation */
c07a80fd 7127
eb160463 7128 *bp++ = (STDCHAR)i; /* store character from PerlIO_getc */
79072805 7129
c07a80fd 7130 if (rslen && (STDCHAR)i == rslast) /* all done for now? */
79072805 7131 goto thats_all_folks;
79072805
LW
7132 }
7133
7134thats_all_folks:
3f7c398e 7135 if ((rslen > 1 && (STRLEN)(bp - (STDCHAR*)SvPVX_const(sv)) < rslen) ||
36477c24 7136 memNE((char*)bp - rslen, rsptr, rslen))
760ac839 7137 goto screamer; /* go back to the fray */
79072805
LW
7138thats_really_all_folks:
7139 if (shortbuffered)
7140 cnt += shortbuffered;
16660edb 7141 DEBUG_P(PerlIO_printf(Perl_debug_log,
1d7c1841 7142 "Screamer: quitting, ptr=%"UVuf", cnt=%ld\n",PTR2UV(ptr),(long)cnt));
cc00df79 7143 PerlIO_set_ptrcnt(fp, (STDCHAR*)ptr, cnt); /* put these back or we're in trouble */
16660edb 7144 DEBUG_P(PerlIO_printf(Perl_debug_log,
1d7c1841 7145 "Screamer: end: FILE * thinks ptr=%"UVuf", cnt=%ld, base=%"UVuf"\n",
1c846c1f 7146 PTR2UV(PerlIO_get_ptr(fp)), (long)PerlIO_get_cnt(fp),
1d7c1841 7147 PTR2UV(PerlIO_has_base (fp) ? PerlIO_get_base(fp) : 0)));
79072805 7148 *bp = '\0';
3f7c398e 7149 SvCUR_set(sv, bp - (STDCHAR*)SvPVX_const(sv)); /* set length */
16660edb 7150 DEBUG_P(PerlIO_printf(Perl_debug_log,
fb73857a 7151 "Screamer: done, len=%ld, string=|%.*s|\n",
3f7c398e 7152 (long)SvCUR(sv),(int)SvCUR(sv),SvPVX_const(sv)));
760ac839
LW
7153 }
7154 else
79072805 7155 {
6edd2cd5 7156 /*The big, slow, and stupid way. */
27da23d5 7157#ifdef USE_HEAP_INSTEAD_OF_STACK /* Even slower way. */
cbbf8932 7158 STDCHAR *buf = NULL;
a02a5408 7159 Newx(buf, 8192, STDCHAR);
6edd2cd5 7160 assert(buf);
4d2c4e07 7161#else
6edd2cd5 7162 STDCHAR buf[8192];
4d2c4e07 7163#endif
79072805 7164
760ac839 7165screamer2:
c07a80fd 7166 if (rslen) {
00b6aa41 7167 register const STDCHAR * const bpe = buf + sizeof(buf);
760ac839 7168 bp = buf;
eb160463 7169 while ((i = PerlIO_getc(fp)) != EOF && (*bp++ = (STDCHAR)i) != rslast && bp < bpe)
760ac839
LW
7170 ; /* keep reading */
7171 cnt = bp - buf;
c07a80fd 7172 }
7173 else {
760ac839 7174 cnt = PerlIO_read(fp,(char*)buf, sizeof(buf));
16660edb 7175 /* Accomodate broken VAXC compiler, which applies U8 cast to
7176 * both args of ?: operator, causing EOF to change into 255
7177 */
37be0adf 7178 if (cnt > 0)
cbe9e203
JH
7179 i = (U8)buf[cnt - 1];
7180 else
37be0adf 7181 i = EOF;
c07a80fd 7182 }
79072805 7183
cbe9e203
JH
7184 if (cnt < 0)
7185 cnt = 0; /* we do need to re-set the sv even when cnt <= 0 */
7186 if (append)
7187 sv_catpvn(sv, (char *) buf, cnt);
7188 else
7189 sv_setpvn(sv, (char *) buf, cnt);
c07a80fd 7190
7191 if (i != EOF && /* joy */
7192 (!rslen ||
7193 SvCUR(sv) < rslen ||
3f7c398e 7194 memNE(SvPVX_const(sv) + SvCUR(sv) - rslen, rsptr, rslen)))
79072805
LW
7195 {
7196 append = -1;
63e4d877
CS
7197 /*
7198 * If we're reading from a TTY and we get a short read,
7199 * indicating that the user hit his EOF character, we need
7200 * to notice it now, because if we try to read from the TTY
7201 * again, the EOF condition will disappear.
7202 *
7203 * The comparison of cnt to sizeof(buf) is an optimization
7204 * that prevents unnecessary calls to feof().
7205 *
7206 * - jik 9/25/96
7207 */
bb7a0f54 7208 if (!(cnt < (I32)sizeof(buf) && PerlIO_eof(fp)))
63e4d877 7209 goto screamer2;
79072805 7210 }
6edd2cd5 7211
27da23d5 7212#ifdef USE_HEAP_INSTEAD_OF_STACK
6edd2cd5
JH
7213 Safefree(buf);
7214#endif
79072805
LW
7215 }
7216
8bfdd7d9 7217 if (rspara) { /* have to do this both before and after */
c07a80fd 7218 while (i != EOF) { /* to make sure file boundaries work right */
760ac839 7219 i = PerlIO_getc(fp);
79072805 7220 if (i != '\n') {
760ac839 7221 PerlIO_ungetc(fp,i);
79072805
LW
7222 break;
7223 }
7224 }
7225 }
c07a80fd 7226
efd8b2ba 7227return_string_or_null:
bd61b366 7228 return (SvCUR(sv) - append) ? SvPVX(sv) : NULL;
79072805
LW
7229}
7230
954c1994
GS
7231/*
7232=for apidoc sv_inc
7233
645c22ef
DM
7234Auto-increment of the value in the SV, doing string to numeric conversion
7235if necessary. Handles 'get' magic.
954c1994
GS
7236
7237=cut
7238*/
7239
79072805 7240void
ac1e9476 7241Perl_sv_inc(pTHX_ register SV *const sv)
79072805 7242{
97aff369 7243 dVAR;
79072805 7244 register char *d;
463ee0b2 7245 int flags;
79072805
LW
7246
7247 if (!sv)
7248 return;
5b295bef 7249 SvGETMAGIC(sv);
ed6116ce 7250 if (SvTHINKFIRST(sv)) {
765f542d
NC
7251 if (SvIsCOW(sv))
7252 sv_force_normal_flags(sv, 0);
0f15f207 7253 if (SvREADONLY(sv)) {
923e4eb5 7254 if (IN_PERL_RUNTIME)
f1f66076 7255 Perl_croak(aTHX_ "%s", PL_no_modify);
0f15f207 7256 }
a0d0e21e 7257 if (SvROK(sv)) {
b5be31e9 7258 IV i;
9e7bc3e8
JD
7259 if (SvAMAGIC(sv) && AMG_CALLun(sv,inc))
7260 return;
56431972 7261 i = PTR2IV(SvRV(sv));
b5be31e9
SM
7262 sv_unref(sv);
7263 sv_setiv(sv, i);
a0d0e21e 7264 }
ed6116ce 7265 }
8990e307 7266 flags = SvFLAGS(sv);
28e5dec8
JH
7267 if ((flags & (SVp_NOK|SVp_IOK)) == SVp_NOK) {
7268 /* It's (privately or publicly) a float, but not tested as an
7269 integer, so test it to see. */
d460ef45 7270 (void) SvIV(sv);
28e5dec8
JH
7271 flags = SvFLAGS(sv);
7272 }
7273 if ((flags & SVf_IOK) || ((flags & (SVp_IOK | SVp_NOK)) == SVp_IOK)) {
7274 /* It's publicly an integer, or privately an integer-not-float */
59d8ce62 7275#ifdef PERL_PRESERVE_IVUV
28e5dec8 7276 oops_its_int:
59d8ce62 7277#endif
25da4f38
IZ
7278 if (SvIsUV(sv)) {
7279 if (SvUVX(sv) == UV_MAX)
a1e868e7 7280 sv_setnv(sv, UV_MAX_P1);
25da4f38
IZ
7281 else
7282 (void)SvIOK_only_UV(sv);
607fa7f2 7283 SvUV_set(sv, SvUVX(sv) + 1);
25da4f38
IZ
7284 } else {
7285 if (SvIVX(sv) == IV_MAX)
28e5dec8 7286 sv_setuv(sv, (UV)IV_MAX + 1);
25da4f38
IZ
7287 else {
7288 (void)SvIOK_only(sv);
45977657 7289 SvIV_set(sv, SvIVX(sv) + 1);
1c846c1f 7290 }
55497cff 7291 }
79072805
LW
7292 return;
7293 }
28e5dec8 7294 if (flags & SVp_NOK) {
b88df990 7295 const NV was = SvNVX(sv);
b68c599a 7296 if (NV_OVERFLOWS_INTEGERS_AT &&
a2a5de95
NC
7297 was >= NV_OVERFLOWS_INTEGERS_AT) {
7298 Perl_ck_warner(aTHX_ packWARN(WARN_IMPRECISION),
7299 "Lost precision when incrementing %" NVff " by 1",
7300 was);
b88df990 7301 }
28e5dec8 7302 (void)SvNOK_only(sv);
b68c599a 7303 SvNV_set(sv, was + 1.0);
28e5dec8
JH
7304 return;
7305 }
7306
3f7c398e 7307 if (!(flags & SVp_POK) || !*SvPVX_const(sv)) {
28e5dec8 7308 if ((flags & SVTYPEMASK) < SVt_PVIV)
f5282e15 7309 sv_upgrade(sv, ((flags & SVTYPEMASK) > SVt_IV ? SVt_PVIV : SVt_IV));
28e5dec8 7310 (void)SvIOK_only(sv);
45977657 7311 SvIV_set(sv, 1);
79072805
LW
7312 return;
7313 }
463ee0b2 7314 d = SvPVX(sv);
79072805
LW
7315 while (isALPHA(*d)) d++;
7316 while (isDIGIT(*d)) d++;
6aff239d 7317 if (d < SvEND(sv)) {
28e5dec8 7318#ifdef PERL_PRESERVE_IVUV
d1be9408 7319 /* Got to punt this as an integer if needs be, but we don't issue
28e5dec8
JH
7320 warnings. Probably ought to make the sv_iv_please() that does
7321 the conversion if possible, and silently. */
504618e9 7322 const int numtype = grok_number(SvPVX_const(sv), SvCUR(sv), NULL);
28e5dec8
JH
7323 if (numtype && !(numtype & IS_NUMBER_INFINITY)) {
7324 /* Need to try really hard to see if it's an integer.
7325 9.22337203685478e+18 is an integer.
7326 but "9.22337203685478e+18" + 0 is UV=9223372036854779904
7327 so $a="9.22337203685478e+18"; $a+0; $a++
7328 needs to be the same as $a="9.22337203685478e+18"; $a++
7329 or we go insane. */
d460ef45 7330
28e5dec8
JH
7331 (void) sv_2iv(sv);
7332 if (SvIOK(sv))
7333 goto oops_its_int;
7334
7335 /* sv_2iv *should* have made this an NV */
7336 if (flags & SVp_NOK) {
7337 (void)SvNOK_only(sv);
9d6ce603 7338 SvNV_set(sv, SvNVX(sv) + 1.0);
28e5dec8
JH
7339 return;
7340 }
7341 /* I don't think we can get here. Maybe I should assert this
7342 And if we do get here I suspect that sv_setnv will croak. NWC
7343 Fall through. */
7344#if defined(USE_LONG_DOUBLE)
7345 DEBUG_c(PerlIO_printf(Perl_debug_log,"sv_inc punt failed to convert '%s' to IOK or NOKp, UV=0x%"UVxf" NV=%"PERL_PRIgldbl"\n",
3f7c398e 7346 SvPVX_const(sv), SvIVX(sv), SvNVX(sv)));
28e5dec8 7347#else
1779d84d 7348 DEBUG_c(PerlIO_printf(Perl_debug_log,"sv_inc punt failed to convert '%s' to IOK or NOKp, UV=0x%"UVxf" NV=%"NVgf"\n",
3f7c398e 7349 SvPVX_const(sv), SvIVX(sv), SvNVX(sv)));
28e5dec8
JH
7350#endif
7351 }
7352#endif /* PERL_PRESERVE_IVUV */
3f7c398e 7353 sv_setnv(sv,Atof(SvPVX_const(sv)) + 1.0);
79072805
LW
7354 return;
7355 }
7356 d--;
3f7c398e 7357 while (d >= SvPVX_const(sv)) {
79072805
LW
7358 if (isDIGIT(*d)) {
7359 if (++*d <= '9')
7360 return;
7361 *(d--) = '0';
7362 }
7363 else {
9d116dd7
JH
7364#ifdef EBCDIC
7365 /* MKS: The original code here died if letters weren't consecutive.
7366 * at least it didn't have to worry about non-C locales. The
7367 * new code assumes that ('z'-'a')==('Z'-'A'), letters are
1c846c1f 7368 * arranged in order (although not consecutively) and that only
9d116dd7
JH
7369 * [A-Za-z] are accepted by isALPHA in the C locale.
7370 */
7371 if (*d != 'z' && *d != 'Z') {
7372 do { ++*d; } while (!isALPHA(*d));
7373 return;
7374 }
7375 *(d--) -= 'z' - 'a';
7376#else
79072805
LW
7377 ++*d;
7378 if (isALPHA(*d))
7379 return;
7380 *(d--) -= 'z' - 'a' + 1;
9d116dd7 7381#endif
79072805
LW
7382 }
7383 }
7384 /* oh,oh, the number grew */
7385 SvGROW(sv, SvCUR(sv) + 2);
b162af07 7386 SvCUR_set(sv, SvCUR(sv) + 1);
3f7c398e 7387 for (d = SvPVX(sv) + SvCUR(sv); d > SvPVX_const(sv); d--)
79072805
LW
7388 *d = d[-1];
7389 if (isDIGIT(d[1]))
7390 *d = '1';
7391 else
7392 *d = d[1];
7393}
7394
954c1994
GS
7395/*
7396=for apidoc sv_dec
7397
645c22ef
DM
7398Auto-decrement of the value in the SV, doing string to numeric conversion
7399if necessary. Handles 'get' magic.
954c1994
GS
7400
7401=cut
7402*/
7403
79072805 7404void
ac1e9476 7405Perl_sv_dec(pTHX_ register SV *const sv)
79072805 7406{
97aff369 7407 dVAR;
463ee0b2
LW
7408 int flags;
7409
79072805
LW
7410 if (!sv)
7411 return;
5b295bef 7412 SvGETMAGIC(sv);
ed6116ce 7413 if (SvTHINKFIRST(sv)) {
765f542d
NC
7414 if (SvIsCOW(sv))
7415 sv_force_normal_flags(sv, 0);
0f15f207 7416 if (SvREADONLY(sv)) {
923e4eb5 7417 if (IN_PERL_RUNTIME)
f1f66076 7418 Perl_croak(aTHX_ "%s", PL_no_modify);
0f15f207 7419 }
a0d0e21e 7420 if (SvROK(sv)) {
b5be31e9 7421 IV i;
9e7bc3e8
JD
7422 if (SvAMAGIC(sv) && AMG_CALLun(sv,dec))
7423 return;
56431972 7424 i = PTR2IV(SvRV(sv));
b5be31e9
SM
7425 sv_unref(sv);
7426 sv_setiv(sv, i);
a0d0e21e 7427 }
ed6116ce 7428 }
28e5dec8
JH
7429 /* Unlike sv_inc we don't have to worry about string-never-numbers
7430 and keeping them magic. But we mustn't warn on punting */
8990e307 7431 flags = SvFLAGS(sv);
28e5dec8
JH
7432 if ((flags & SVf_IOK) || ((flags & (SVp_IOK | SVp_NOK)) == SVp_IOK)) {
7433 /* It's publicly an integer, or privately an integer-not-float */
59d8ce62 7434#ifdef PERL_PRESERVE_IVUV
28e5dec8 7435 oops_its_int:
59d8ce62 7436#endif
25da4f38
IZ
7437 if (SvIsUV(sv)) {
7438 if (SvUVX(sv) == 0) {
7439 (void)SvIOK_only(sv);
45977657 7440 SvIV_set(sv, -1);
25da4f38
IZ
7441 }
7442 else {
7443 (void)SvIOK_only_UV(sv);
f4eee32f 7444 SvUV_set(sv, SvUVX(sv) - 1);
1c846c1f 7445 }
25da4f38 7446 } else {
b88df990
NC
7447 if (SvIVX(sv) == IV_MIN) {
7448 sv_setnv(sv, (NV)IV_MIN);
7449 goto oops_its_num;
7450 }
25da4f38
IZ
7451 else {
7452 (void)SvIOK_only(sv);
45977657 7453 SvIV_set(sv, SvIVX(sv) - 1);
1c846c1f 7454 }
55497cff 7455 }
7456 return;
7457 }
28e5dec8 7458 if (flags & SVp_NOK) {
b88df990
NC
7459 oops_its_num:
7460 {
7461 const NV was = SvNVX(sv);
b68c599a 7462 if (NV_OVERFLOWS_INTEGERS_AT &&
a2a5de95
NC
7463 was <= -NV_OVERFLOWS_INTEGERS_AT) {
7464 Perl_ck_warner(aTHX_ packWARN(WARN_IMPRECISION),
7465 "Lost precision when decrementing %" NVff " by 1",
7466 was);
b88df990
NC
7467 }
7468 (void)SvNOK_only(sv);
b68c599a 7469 SvNV_set(sv, was - 1.0);
b88df990
NC
7470 return;
7471 }
28e5dec8 7472 }
8990e307 7473 if (!(flags & SVp_POK)) {
ef088171
NC
7474 if ((flags & SVTYPEMASK) < SVt_PVIV)
7475 sv_upgrade(sv, ((flags & SVTYPEMASK) > SVt_IV) ? SVt_PVIV : SVt_IV);
7476 SvIV_set(sv, -1);
7477 (void)SvIOK_only(sv);
79072805
LW
7478 return;
7479 }
28e5dec8
JH
7480#ifdef PERL_PRESERVE_IVUV
7481 {
504618e9 7482 const int numtype = grok_number(SvPVX_const(sv), SvCUR(sv), NULL);
28e5dec8
JH
7483 if (numtype && !(numtype & IS_NUMBER_INFINITY)) {
7484 /* Need to try really hard to see if it's an integer.
7485 9.22337203685478e+18 is an integer.
7486 but "9.22337203685478e+18" + 0 is UV=9223372036854779904
7487 so $a="9.22337203685478e+18"; $a+0; $a--
7488 needs to be the same as $a="9.22337203685478e+18"; $a--
7489 or we go insane. */
d460ef45 7490
28e5dec8
JH
7491 (void) sv_2iv(sv);
7492 if (SvIOK(sv))
7493 goto oops_its_int;
7494
7495 /* sv_2iv *should* have made this an NV */
7496 if (flags & SVp_NOK) {
7497 (void)SvNOK_only(sv);
9d6ce603 7498 SvNV_set(sv, SvNVX(sv) - 1.0);
28e5dec8
JH
7499 return;
7500 }
7501 /* I don't think we can get here. Maybe I should assert this
7502 And if we do get here I suspect that sv_setnv will croak. NWC
7503 Fall through. */
7504#if defined(USE_LONG_DOUBLE)
7505 DEBUG_c(PerlIO_printf(Perl_debug_log,"sv_dec punt failed to convert '%s' to IOK or NOKp, UV=0x%"UVxf" NV=%"PERL_PRIgldbl"\n",
3f7c398e 7506 SvPVX_const(sv), SvIVX(sv), SvNVX(sv)));
28e5dec8 7507#else
1779d84d 7508 DEBUG_c(PerlIO_printf(Perl_debug_log,"sv_dec punt failed to convert '%s' to IOK or NOKp, UV=0x%"UVxf" NV=%"NVgf"\n",
3f7c398e 7509 SvPVX_const(sv), SvIVX(sv), SvNVX(sv)));
28e5dec8
JH
7510#endif
7511 }
7512 }
7513#endif /* PERL_PRESERVE_IVUV */
3f7c398e 7514 sv_setnv(sv,Atof(SvPVX_const(sv)) - 1.0); /* punt */
79072805
LW
7515}
7516
81041c50
YO
7517/* this define is used to eliminate a chunk of duplicated but shared logic
7518 * it has the suffix __SV_C to signal that it isnt API, and isnt meant to be
7519 * used anywhere but here - yves
7520 */
7521#define PUSH_EXTEND_MORTAL__SV_C(AnSv) \
7522 STMT_START { \
7523 EXTEND_MORTAL(1); \
7524 PL_tmps_stack[++PL_tmps_ix] = (AnSv); \
7525 } STMT_END
7526
954c1994
GS
7527/*
7528=for apidoc sv_mortalcopy
7529
645c22ef 7530Creates a new SV which is a copy of the original SV (using C<sv_setsv>).
d4236ebc
DM
7531The new SV is marked as mortal. It will be destroyed "soon", either by an
7532explicit call to FREETMPS, or by an implicit call at places such as
7533statement boundaries. See also C<sv_newmortal> and C<sv_2mortal>.
954c1994
GS
7534
7535=cut
7536*/
7537
79072805
LW
7538/* Make a string that will exist for the duration of the expression
7539 * evaluation. Actually, it may have to last longer than that, but
7540 * hopefully we won't free it until it has been assigned to a
7541 * permanent location. */
7542
7543SV *
ac1e9476 7544Perl_sv_mortalcopy(pTHX_ SV *const oldstr)
79072805 7545{
97aff369 7546 dVAR;
463ee0b2 7547 register SV *sv;
b881518d 7548
4561caa4 7549 new_SV(sv);
79072805 7550 sv_setsv(sv,oldstr);
81041c50 7551 PUSH_EXTEND_MORTAL__SV_C(sv);
8990e307
LW
7552 SvTEMP_on(sv);
7553 return sv;
7554}
7555
954c1994
GS
7556/*
7557=for apidoc sv_newmortal
7558
645c22ef 7559Creates a new null SV which is mortal. The reference count of the SV is
d4236ebc
DM
7560set to 1. It will be destroyed "soon", either by an explicit call to
7561FREETMPS, or by an implicit call at places such as statement boundaries.
7562See also C<sv_mortalcopy> and C<sv_2mortal>.
954c1994
GS
7563
7564=cut
7565*/
7566
8990e307 7567SV *
864dbfa3 7568Perl_sv_newmortal(pTHX)
8990e307 7569{
97aff369 7570 dVAR;
8990e307
LW
7571 register SV *sv;
7572
4561caa4 7573 new_SV(sv);
8990e307 7574 SvFLAGS(sv) = SVs_TEMP;
81041c50 7575 PUSH_EXTEND_MORTAL__SV_C(sv);
79072805
LW
7576 return sv;
7577}
7578
59cd0e26
NC
7579
7580/*
7581=for apidoc newSVpvn_flags
7582
7583Creates a new SV and copies a string into it. The reference count for the
7584SV is set to 1. Note that if C<len> is zero, Perl will create a zero length
7585string. You are responsible for ensuring that the source string is at least
7586C<len> bytes long. If the C<s> argument is NULL the new SV will be undefined.
7587Currently the only flag bits accepted are C<SVf_UTF8> and C<SVs_TEMP>.
7588If C<SVs_TEMP> is set, then C<sv2mortal()> is called on the result before
7589returning. If C<SVf_UTF8> is set, then it will be set on the new SV.
7590C<newSVpvn_utf8()> is a convenience wrapper for this function, defined as
7591
7592 #define newSVpvn_utf8(s, len, u) \
7593 newSVpvn_flags((s), (len), (u) ? SVf_UTF8 : 0)
7594
7595=cut
7596*/
7597
7598SV *
23f13727 7599Perl_newSVpvn_flags(pTHX_ const char *const s, const STRLEN len, const U32 flags)
59cd0e26
NC
7600{
7601 dVAR;
7602 register SV *sv;
7603
7604 /* All the flags we don't support must be zero.
7605 And we're new code so I'm going to assert this from the start. */
7606 assert(!(flags & ~(SVf_UTF8|SVs_TEMP)));
7607 new_SV(sv);
7608 sv_setpvn(sv,s,len);
d21488d7
YO
7609
7610 /* This code used to a sv_2mortal(), however we now unroll the call to sv_2mortal()
7611 * and do what it does outselves here.
7612 * Since we have asserted that flags can only have the SVf_UTF8 and/or SVs_TEMP flags
7613 * set above we can use it to enable the sv flags directly (bypassing SvTEMP_on), which
7614 * in turn means we dont need to mask out the SVf_UTF8 flag below, which means that we
7615 * eleminate quite a few steps than it looks - Yves (explaining patch by gfx)
7616 */
7617
6dfeccca
GF
7618 SvFLAGS(sv) |= flags;
7619
7620 if(flags & SVs_TEMP){
81041c50 7621 PUSH_EXTEND_MORTAL__SV_C(sv);
6dfeccca
GF
7622 }
7623
7624 return sv;
59cd0e26
NC
7625}
7626
954c1994
GS
7627/*
7628=for apidoc sv_2mortal
7629
d4236ebc
DM
7630Marks an existing SV as mortal. The SV will be destroyed "soon", either
7631by an explicit call to FREETMPS, or by an implicit call at places such as
37d2ac18
NC
7632statement boundaries. SvTEMP() is turned on which means that the SV's
7633string buffer can be "stolen" if this SV is copied. See also C<sv_newmortal>
7634and C<sv_mortalcopy>.
954c1994
GS
7635
7636=cut
7637*/
7638
79072805 7639SV *
23f13727 7640Perl_sv_2mortal(pTHX_ register SV *const sv)
79072805 7641{
27da23d5 7642 dVAR;
79072805 7643 if (!sv)
7a5b473e 7644 return NULL;
d689ffdd 7645 if (SvREADONLY(sv) && SvIMMORTAL(sv))
11162842 7646 return sv;
81041c50 7647 PUSH_EXTEND_MORTAL__SV_C(sv);
8990e307 7648 SvTEMP_on(sv);
79072805
LW
7649 return sv;
7650}
7651
954c1994
GS
7652/*
7653=for apidoc newSVpv
7654
7655Creates a new SV and copies a string into it. The reference count for the
7656SV is set to 1. If C<len> is zero, Perl will compute the length using
7657strlen(). For efficiency, consider using C<newSVpvn> instead.
7658
7659=cut
7660*/
7661
79072805 7662SV *
23f13727 7663Perl_newSVpv(pTHX_ const char *const s, const STRLEN len)
79072805 7664{
97aff369 7665 dVAR;
463ee0b2 7666 register SV *sv;
79072805 7667
4561caa4 7668 new_SV(sv);
ddfa59c7 7669 sv_setpvn(sv, s, len || s == NULL ? len : strlen(s));
79072805
LW
7670 return sv;
7671}
7672
954c1994
GS
7673/*
7674=for apidoc newSVpvn
7675
7676Creates a new SV and copies a string into it. The reference count for the
1c846c1f 7677SV is set to 1. Note that if C<len> is zero, Perl will create a zero length
954c1994 7678string. You are responsible for ensuring that the source string is at least
9e09f5f2 7679C<len> bytes long. If the C<s> argument is NULL the new SV will be undefined.
954c1994
GS
7680
7681=cut
7682*/
7683
9da1e3b5 7684SV *
23f13727 7685Perl_newSVpvn(pTHX_ const char *const s, const STRLEN len)
9da1e3b5 7686{
97aff369 7687 dVAR;
9da1e3b5
MUN
7688 register SV *sv;
7689
7690 new_SV(sv);
9da1e3b5
MUN
7691 sv_setpvn(sv,s,len);
7692 return sv;
7693}
7694
740cce10 7695/*
926f8064 7696=for apidoc newSVhek
bd08039b
NC
7697
7698Creates a new SV from the hash key structure. It will generate scalars that
5aaec2b4
NC
7699point to the shared string table where possible. Returns a new (undefined)
7700SV if the hek is NULL.
bd08039b
NC
7701
7702=cut
7703*/
7704
7705SV *
23f13727 7706Perl_newSVhek(pTHX_ const HEK *const hek)
bd08039b 7707{
97aff369 7708 dVAR;
5aaec2b4
NC
7709 if (!hek) {
7710 SV *sv;
7711
7712 new_SV(sv);
7713 return sv;
7714 }
7715
bd08039b
NC
7716 if (HEK_LEN(hek) == HEf_SVKEY) {
7717 return newSVsv(*(SV**)HEK_KEY(hek));
7718 } else {
7719 const int flags = HEK_FLAGS(hek);
7720 if (flags & HVhek_WASUTF8) {
7721 /* Trouble :-)
7722 Andreas would like keys he put in as utf8 to come back as utf8
7723 */
7724 STRLEN utf8_len = HEK_LEN(hek);
b64e5050
AL
7725 const U8 *as_utf8 = bytes_to_utf8 ((U8*)HEK_KEY(hek), &utf8_len);
7726 SV * const sv = newSVpvn ((const char*)as_utf8, utf8_len);
bd08039b
NC
7727
7728 SvUTF8_on (sv);
7729 Safefree (as_utf8); /* bytes_to_utf8() allocates a new string */
7730 return sv;
45e34800 7731 } else if (flags & (HVhek_REHASH|HVhek_UNSHARED)) {
bd08039b
NC
7732 /* We don't have a pointer to the hv, so we have to replicate the
7733 flag into every HEK. This hv is using custom a hasing
7734 algorithm. Hence we can't return a shared string scalar, as
7735 that would contain the (wrong) hash value, and might get passed
45e34800
NC
7736 into an hv routine with a regular hash.
7737 Similarly, a hash that isn't using shared hash keys has to have
7738 the flag in every key so that we know not to try to call
7739 share_hek_kek on it. */
bd08039b 7740
b64e5050 7741 SV * const sv = newSVpvn (HEK_KEY(hek), HEK_LEN(hek));
bd08039b
NC
7742 if (HEK_UTF8(hek))
7743 SvUTF8_on (sv);
7744 return sv;
7745 }
7746 /* This will be overwhelminly the most common case. */
409dfe77
NC
7747 {
7748 /* Inline most of newSVpvn_share(), because share_hek_hek() is far
7749 more efficient than sharepvn(). */
7750 SV *sv;
7751
7752 new_SV(sv);
7753 sv_upgrade(sv, SVt_PV);
7754 SvPV_set(sv, (char *)HEK_KEY(share_hek_hek(hek)));
7755 SvCUR_set(sv, HEK_LEN(hek));
7756 SvLEN_set(sv, 0);
7757 SvREADONLY_on(sv);
7758 SvFAKE_on(sv);
7759 SvPOK_on(sv);
7760 if (HEK_UTF8(hek))
7761 SvUTF8_on(sv);
7762 return sv;
7763 }
bd08039b
NC
7764 }
7765}
7766
1c846c1f
NIS
7767/*
7768=for apidoc newSVpvn_share
7769
3f7c398e 7770Creates a new SV with its SvPVX_const pointing to a shared string in the string
645c22ef 7771table. If the string does not already exist in the table, it is created
758fcfc1
VP
7772first. Turns on READONLY and FAKE. If the C<hash> parameter is non-zero, that
7773value is used; otherwise the hash is computed. The string's hash can be later
7774be retrieved from the SV with the C<SvSHARED_HASH()> macro. The idea here is
7775that as the string table is used for shared hash keys these strings will have
7776SvPVX_const == HeKEY and hash lookup will avoid string compare.
1c846c1f
NIS
7777
7778=cut
7779*/
7780
7781SV *
c3654f1a 7782Perl_newSVpvn_share(pTHX_ const char *src, I32 len, U32 hash)
1c846c1f 7783{
97aff369 7784 dVAR;
1c846c1f 7785 register SV *sv;
c3654f1a 7786 bool is_utf8 = FALSE;
a51caccf
NC
7787 const char *const orig_src = src;
7788
c3654f1a 7789 if (len < 0) {
77caf834 7790 STRLEN tmplen = -len;
c3654f1a 7791 is_utf8 = TRUE;
75a54232 7792 /* See the note in hv.c:hv_fetch() --jhi */
e1ec3a88 7793 src = (char*)bytes_from_utf8((const U8*)src, &tmplen, &is_utf8);
75a54232
JH
7794 len = tmplen;
7795 }
1c846c1f 7796 if (!hash)
5afd6d42 7797 PERL_HASH(hash, src, len);
1c846c1f 7798 new_SV(sv);
f46ee248
NC
7799 /* The logic for this is inlined in S_mro_get_linear_isa_dfs(), so if it
7800 changes here, update it there too. */
bdd68bc3 7801 sv_upgrade(sv, SVt_PV);
f880fe2f 7802 SvPV_set(sv, sharepvn(src, is_utf8?-len:len, hash));
b162af07 7803 SvCUR_set(sv, len);
b162af07 7804 SvLEN_set(sv, 0);
1c846c1f
NIS
7805 SvREADONLY_on(sv);
7806 SvFAKE_on(sv);
7807 SvPOK_on(sv);
c3654f1a
IH
7808 if (is_utf8)
7809 SvUTF8_on(sv);
a51caccf
NC
7810 if (src != orig_src)
7811 Safefree(src);
1c846c1f
NIS
7812 return sv;
7813}
7814
645c22ef 7815
cea2e8a9 7816#if defined(PERL_IMPLICIT_CONTEXT)
645c22ef
DM
7817
7818/* pTHX_ magic can't cope with varargs, so this is a no-context
7819 * version of the main function, (which may itself be aliased to us).
7820 * Don't access this version directly.
7821 */
7822
46fc3d4c 7823SV *
23f13727 7824Perl_newSVpvf_nocontext(const char *const pat, ...)
46fc3d4c 7825{
cea2e8a9 7826 dTHX;
46fc3d4c 7827 register SV *sv;
7828 va_list args;
7918f24d
NC
7829
7830 PERL_ARGS_ASSERT_NEWSVPVF_NOCONTEXT;
7831
46fc3d4c 7832 va_start(args, pat);
c5be433b 7833 sv = vnewSVpvf(pat, &args);
46fc3d4c 7834 va_end(args);
7835 return sv;
7836}
cea2e8a9 7837#endif
46fc3d4c 7838
954c1994
GS
7839/*
7840=for apidoc newSVpvf
7841
645c22ef 7842Creates a new SV and initializes it with the string formatted like
954c1994
GS
7843C<sprintf>.
7844
7845=cut
7846*/
7847
cea2e8a9 7848SV *
23f13727 7849Perl_newSVpvf(pTHX_ const char *const pat, ...)
cea2e8a9
GS
7850{
7851 register SV *sv;
7852 va_list args;
7918f24d
NC
7853
7854 PERL_ARGS_ASSERT_NEWSVPVF;
7855
cea2e8a9 7856 va_start(args, pat);
c5be433b 7857 sv = vnewSVpvf(pat, &args);
cea2e8a9
GS
7858 va_end(args);
7859 return sv;
7860}
46fc3d4c 7861
645c22ef
DM
7862/* backend for newSVpvf() and newSVpvf_nocontext() */
7863
79072805 7864SV *
23f13727 7865Perl_vnewSVpvf(pTHX_ const char *const pat, va_list *const args)
c5be433b 7866{
97aff369 7867 dVAR;
c5be433b 7868 register SV *sv;
7918f24d
NC
7869
7870 PERL_ARGS_ASSERT_VNEWSVPVF;
7871
c5be433b 7872 new_SV(sv);
4608196e 7873 sv_vsetpvfn(sv, pat, strlen(pat), args, NULL, 0, NULL);
c5be433b
GS
7874 return sv;
7875}
7876
954c1994
GS
7877/*
7878=for apidoc newSVnv
7879
7880Creates a new SV and copies a floating point value into it.
7881The reference count for the SV is set to 1.
7882
7883=cut
7884*/
7885
c5be433b 7886SV *
23f13727 7887Perl_newSVnv(pTHX_ const NV n)
79072805 7888{
97aff369 7889 dVAR;
463ee0b2 7890 register SV *sv;
79072805 7891
4561caa4 7892 new_SV(sv);
79072805
LW
7893 sv_setnv(sv,n);
7894 return sv;
7895}
7896
954c1994
GS
7897/*
7898=for apidoc newSViv
7899
7900Creates a new SV and copies an integer into it. The reference count for the
7901SV is set to 1.
7902
7903=cut
7904*/
7905
79072805 7906SV *
23f13727 7907Perl_newSViv(pTHX_ const IV i)
79072805 7908{
97aff369 7909 dVAR;
463ee0b2 7910 register SV *sv;
79072805 7911
4561caa4 7912 new_SV(sv);
79072805
LW
7913 sv_setiv(sv,i);
7914 return sv;
7915}
7916
954c1994 7917/*
1a3327fb
JH
7918=for apidoc newSVuv
7919
7920Creates a new SV and copies an unsigned integer into it.
7921The reference count for the SV is set to 1.
7922
7923=cut
7924*/
7925
7926SV *
23f13727 7927Perl_newSVuv(pTHX_ const UV u)
1a3327fb 7928{
97aff369 7929 dVAR;
1a3327fb
JH
7930 register SV *sv;
7931
7932 new_SV(sv);
7933 sv_setuv(sv,u);
7934 return sv;
7935}
7936
7937/*
b9f83d2f
NC
7938=for apidoc newSV_type
7939
c41f7ed2 7940Creates a new SV, of the type specified. The reference count for the new SV
b9f83d2f
NC
7941is set to 1.
7942
7943=cut
7944*/
7945
7946SV *
fe9845cc 7947Perl_newSV_type(pTHX_ const svtype type)
b9f83d2f
NC
7948{
7949 register SV *sv;
7950
7951 new_SV(sv);
7952 sv_upgrade(sv, type);
7953 return sv;
7954}
7955
7956/*
954c1994
GS
7957=for apidoc newRV_noinc
7958
7959Creates an RV wrapper for an SV. The reference count for the original
7960SV is B<not> incremented.
7961
7962=cut
7963*/
7964
2304df62 7965SV *
23f13727 7966Perl_newRV_noinc(pTHX_ SV *const tmpRef)
2304df62 7967{
97aff369 7968 dVAR;
4df7f6af 7969 register SV *sv = newSV_type(SVt_IV);
7918f24d
NC
7970
7971 PERL_ARGS_ASSERT_NEWRV_NOINC;
7972
76e3520e 7973 SvTEMP_off(tmpRef);
b162af07 7974 SvRV_set(sv, tmpRef);
2304df62 7975 SvROK_on(sv);
2304df62
AD
7976 return sv;
7977}
7978
ff276b08 7979/* newRV_inc is the official function name to use now.
645c22ef
DM
7980 * newRV_inc is in fact #defined to newRV in sv.h
7981 */
7982
5f05dabc 7983SV *
23f13727 7984Perl_newRV(pTHX_ SV *const sv)
5f05dabc 7985{
97aff369 7986 dVAR;
7918f24d
NC
7987
7988 PERL_ARGS_ASSERT_NEWRV;
7989
7f466ec7 7990 return newRV_noinc(SvREFCNT_inc_simple_NN(sv));
5f05dabc 7991}
5f05dabc 7992
954c1994
GS
7993/*
7994=for apidoc newSVsv
7995
7996Creates a new SV which is an exact duplicate of the original SV.
645c22ef 7997(Uses C<sv_setsv>).
954c1994
GS
7998
7999=cut
8000*/
8001
79072805 8002SV *
23f13727 8003Perl_newSVsv(pTHX_ register SV *const old)
79072805 8004{
97aff369 8005 dVAR;
463ee0b2 8006 register SV *sv;
79072805
LW
8007
8008 if (!old)
7a5b473e 8009 return NULL;
8990e307 8010 if (SvTYPE(old) == SVTYPEMASK) {
9b387841 8011 Perl_ck_warner_d(aTHX_ packWARN(WARN_INTERNAL), "semi-panic: attempt to dup freed string");
a0714e2c 8012 return NULL;
79072805 8013 }
4561caa4 8014 new_SV(sv);
e90aabeb
NC
8015 /* SV_GMAGIC is the default for sv_setv()
8016 SV_NOSTEAL prevents TEMP buffers being, well, stolen, and saves games
8017 with SvTEMP_off and SvTEMP_on round a call to sv_setsv. */
8018 sv_setsv_flags(sv, old, SV_GMAGIC | SV_NOSTEAL);
463ee0b2 8019 return sv;
79072805
LW
8020}
8021
645c22ef
DM
8022/*
8023=for apidoc sv_reset
8024
8025Underlying implementation for the C<reset> Perl function.
8026Note that the perl-level function is vaguely deprecated.
8027
8028=cut
8029*/
8030
79072805 8031void
23f13727 8032Perl_sv_reset(pTHX_ register const char *s, HV *const stash)
79072805 8033{
27da23d5 8034 dVAR;
4802d5d7 8035 char todo[PERL_UCHAR_MAX+1];
79072805 8036
7918f24d
NC
8037 PERL_ARGS_ASSERT_SV_RESET;
8038
49d8d3a1
MB
8039 if (!stash)
8040 return;
8041
79072805 8042 if (!*s) { /* reset ?? searches */
daba3364 8043 MAGIC * const mg = mg_find((const SV *)stash, PERL_MAGIC_symtab);
8d2f4536 8044 if (mg) {
c2b1997a
NC
8045 const U32 count = mg->mg_len / sizeof(PMOP**);
8046 PMOP **pmp = (PMOP**) mg->mg_ptr;
8047 PMOP *const *const end = pmp + count;
8048
8049 while (pmp < end) {
c737faaf 8050#ifdef USE_ITHREADS
c2b1997a 8051 SvREADONLY_off(PL_regex_pad[(*pmp)->op_pmoffset]);
c737faaf 8052#else
c2b1997a 8053 (*pmp)->op_pmflags &= ~PMf_USED;
c737faaf 8054#endif
c2b1997a 8055 ++pmp;
8d2f4536 8056 }
79072805
LW
8057 }
8058 return;
8059 }
8060
8061 /* reset variables */
8062
8063 if (!HvARRAY(stash))
8064 return;
463ee0b2
LW
8065
8066 Zero(todo, 256, char);
79072805 8067 while (*s) {
b464bac0
AL
8068 I32 max;
8069 I32 i = (unsigned char)*s;
79072805
LW
8070 if (s[1] == '-') {
8071 s += 2;
8072 }
4802d5d7 8073 max = (unsigned char)*s++;
79072805 8074 for ( ; i <= max; i++) {
463ee0b2
LW
8075 todo[i] = 1;
8076 }
a0d0e21e 8077 for (i = 0; i <= (I32) HvMAX(stash); i++) {
b464bac0 8078 HE *entry;
79072805 8079 for (entry = HvARRAY(stash)[i];
9e35f4b3
GS
8080 entry;
8081 entry = HeNEXT(entry))
8082 {
b464bac0
AL
8083 register GV *gv;
8084 register SV *sv;
8085
1edc1566 8086 if (!todo[(U8)*HeKEY(entry)])
463ee0b2 8087 continue;
159b6efe 8088 gv = MUTABLE_GV(HeVAL(entry));
79072805 8089 sv = GvSV(gv);
e203899d
NC
8090 if (sv) {
8091 if (SvTHINKFIRST(sv)) {
8092 if (!SvREADONLY(sv) && SvROK(sv))
8093 sv_unref(sv);
8094 /* XXX Is this continue a bug? Why should THINKFIRST
8095 exempt us from resetting arrays and hashes? */
8096 continue;
8097 }
8098 SvOK_off(sv);
8099 if (SvTYPE(sv) >= SVt_PV) {
8100 SvCUR_set(sv, 0);
bd61b366 8101 if (SvPVX_const(sv) != NULL)
e203899d
NC
8102 *SvPVX(sv) = '\0';
8103 SvTAINT(sv);
8104 }
79072805
LW
8105 }
8106 if (GvAV(gv)) {
8107 av_clear(GvAV(gv));
8108 }
bfcb3514 8109 if (GvHV(gv) && !HvNAME_get(GvHV(gv))) {
b0269e46
AB
8110#if defined(VMS)
8111 Perl_die(aTHX_ "Can't reset %%ENV on this system");
8112#else /* ! VMS */
463ee0b2 8113 hv_clear(GvHV(gv));
b0269e46
AB
8114# if defined(USE_ENVIRON_ARRAY)
8115 if (gv == PL_envgv)
8116 my_clearenv();
8117# endif /* USE_ENVIRON_ARRAY */
8118#endif /* VMS */
79072805
LW
8119 }
8120 }
8121 }
8122 }
8123}
8124
645c22ef
DM
8125/*
8126=for apidoc sv_2io
8127
8128Using various gambits, try to get an IO from an SV: the IO slot if its a
8129GV; or the recursive result if we're an RV; or the IO slot of the symbol
8130named after the PV if we're a string.
8131
8132=cut
8133*/
8134
46fc3d4c 8135IO*
23f13727 8136Perl_sv_2io(pTHX_ SV *const sv)
46fc3d4c 8137{
8138 IO* io;
8139 GV* gv;
8140
7918f24d
NC
8141 PERL_ARGS_ASSERT_SV_2IO;
8142
46fc3d4c 8143 switch (SvTYPE(sv)) {
8144 case SVt_PVIO:
a45c7426 8145 io = MUTABLE_IO(sv);
46fc3d4c 8146 break;
8147 case SVt_PVGV:
6e592b3a 8148 if (isGV_with_GP(sv)) {
159b6efe 8149 gv = MUTABLE_GV(sv);
6e592b3a
BM
8150 io = GvIO(gv);
8151 if (!io)
8152 Perl_croak(aTHX_ "Bad filehandle: %s", GvNAME(gv));
8153 break;
8154 }
8155 /* FALL THROUGH */
46fc3d4c 8156 default:
8157 if (!SvOK(sv))
cea2e8a9 8158 Perl_croak(aTHX_ PL_no_usym, "filehandle");
46fc3d4c 8159 if (SvROK(sv))
8160 return sv_2io(SvRV(sv));
f776e3cd 8161 gv = gv_fetchsv(sv, 0, SVt_PVIO);
46fc3d4c 8162 if (gv)
8163 io = GvIO(gv);
8164 else
8165 io = 0;
8166 if (!io)
be2597df 8167 Perl_croak(aTHX_ "Bad filehandle: %"SVf, SVfARG(sv));
46fc3d4c 8168 break;
8169 }
8170 return io;
8171}
8172
645c22ef
DM
8173/*
8174=for apidoc sv_2cv
8175
8176Using various gambits, try to get a CV from an SV; in addition, try if
8177possible to set C<*st> and C<*gvp> to the stash and GV associated with it.
8e324704 8178The flags in C<lref> are passed to gv_fetchsv.
645c22ef
DM
8179
8180=cut
8181*/
8182
79072805 8183CV *
23f13727 8184Perl_sv_2cv(pTHX_ SV *sv, HV **const st, GV **const gvp, const I32 lref)
79072805 8185{
27da23d5 8186 dVAR;
a0714e2c 8187 GV *gv = NULL;
601f1833 8188 CV *cv = NULL;
79072805 8189
7918f24d
NC
8190 PERL_ARGS_ASSERT_SV_2CV;
8191
85dec29a
NC
8192 if (!sv) {
8193 *st = NULL;
8194 *gvp = NULL;
8195 return NULL;
8196 }
79072805 8197 switch (SvTYPE(sv)) {
79072805
LW
8198 case SVt_PVCV:
8199 *st = CvSTASH(sv);
a0714e2c 8200 *gvp = NULL;
ea726b52 8201 return MUTABLE_CV(sv);
79072805
LW
8202 case SVt_PVHV:
8203 case SVt_PVAV:
ef58ba18 8204 *st = NULL;
a0714e2c 8205 *gvp = NULL;
601f1833 8206 return NULL;
8990e307 8207 case SVt_PVGV:
6e592b3a 8208 if (isGV_with_GP(sv)) {
159b6efe 8209 gv = MUTABLE_GV(sv);
6e592b3a
BM
8210 *gvp = gv;
8211 *st = GvESTASH(gv);
8212 goto fix_gv;
8213 }
8214 /* FALL THROUGH */
8990e307 8215
79072805 8216 default:
a0d0e21e 8217 if (SvROK(sv)) {
823a54a3 8218 SV * const *sp = &sv; /* Used in tryAMAGICunDEREF macro. */
c4f3bd1e 8219 SvGETMAGIC(sv);
f5284f61
IZ
8220 tryAMAGICunDEREF(to_cv);
8221
62f274bf
GS
8222 sv = SvRV(sv);
8223 if (SvTYPE(sv) == SVt_PVCV) {
ea726b52 8224 cv = MUTABLE_CV(sv);
a0714e2c 8225 *gvp = NULL;
62f274bf
GS
8226 *st = CvSTASH(cv);
8227 return cv;
8228 }
6e592b3a 8229 else if(isGV_with_GP(sv))
159b6efe 8230 gv = MUTABLE_GV(sv);
62f274bf 8231 else
cea2e8a9 8232 Perl_croak(aTHX_ "Not a subroutine reference");
a0d0e21e 8233 }
6e592b3a 8234 else if (isGV_with_GP(sv)) {
9d0f7ed7 8235 SvGETMAGIC(sv);
159b6efe 8236 gv = MUTABLE_GV(sv);
9d0f7ed7 8237 }
79072805 8238 else
9d0f7ed7 8239 gv = gv_fetchsv(sv, lref, SVt_PVCV); /* Calls get magic */
79072805 8240 *gvp = gv;
ef58ba18
NC
8241 if (!gv) {
8242 *st = NULL;
601f1833 8243 return NULL;
ef58ba18 8244 }
e26df76a 8245 /* Some flags to gv_fetchsv mean don't really create the GV */
6e592b3a 8246 if (!isGV_with_GP(gv)) {
e26df76a
NC
8247 *st = NULL;
8248 return NULL;
8249 }
79072805 8250 *st = GvESTASH(gv);
8990e307 8251 fix_gv:
8ebc5c01 8252 if (lref && !GvCVu(gv)) {
4633a7c4 8253 SV *tmpsv;
748a9306 8254 ENTER;
561b68a9 8255 tmpsv = newSV(0);
bd61b366 8256 gv_efullname3(tmpsv, gv, NULL);
f6ec51f7
GS
8257 /* XXX this is probably not what they think they're getting.
8258 * It has the same effect as "sub name;", i.e. just a forward
8259 * declaration! */
774d564b 8260 newSUB(start_subparse(FALSE, 0),
4633a7c4 8261 newSVOP(OP_CONST, 0, tmpsv),
5f66b61c 8262 NULL, NULL);
748a9306 8263 LEAVE;
8ebc5c01 8264 if (!GvCVu(gv))
35c1215d 8265 Perl_croak(aTHX_ "Unable to create sub named \"%"SVf"\"",
4052d21c 8266 SVfARG(SvOK(sv) ? sv : &PL_sv_no));
8990e307 8267 }
8ebc5c01 8268 return GvCVu(gv);
79072805
LW
8269 }
8270}
8271
c461cf8f
JH
8272/*
8273=for apidoc sv_true
8274
8275Returns true if the SV has a true value by Perl's rules.
645c22ef
DM
8276Use the C<SvTRUE> macro instead, which may call C<sv_true()> or may
8277instead use an in-line version.
c461cf8f
JH
8278
8279=cut
8280*/
8281
79072805 8282I32
23f13727 8283Perl_sv_true(pTHX_ register SV *const sv)
79072805 8284{
8990e307
LW
8285 if (!sv)
8286 return 0;
79072805 8287 if (SvPOK(sv)) {
823a54a3
AL
8288 register const XPV* const tXpv = (XPV*)SvANY(sv);
8289 if (tXpv &&
c2f1de04 8290 (tXpv->xpv_cur > 1 ||
339049b0 8291 (tXpv->xpv_cur && *sv->sv_u.svu_pv != '0')))
79072805
LW
8292 return 1;
8293 else
8294 return 0;
8295 }
8296 else {
8297 if (SvIOK(sv))
463ee0b2 8298 return SvIVX(sv) != 0;
79072805
LW
8299 else {
8300 if (SvNOK(sv))
463ee0b2 8301 return SvNVX(sv) != 0.0;
79072805 8302 else
463ee0b2 8303 return sv_2bool(sv);
79072805
LW
8304 }
8305 }
8306}
79072805 8307
645c22ef 8308/*
c461cf8f
JH
8309=for apidoc sv_pvn_force
8310
8311Get a sensible string out of the SV somehow.
645c22ef
DM
8312A private implementation of the C<SvPV_force> macro for compilers which
8313can't cope with complex macro expressions. Always use the macro instead.
c461cf8f 8314
8d6d96c1
HS
8315=for apidoc sv_pvn_force_flags
8316
8317Get a sensible string out of the SV somehow.
8318If C<flags> has C<SV_GMAGIC> bit set, will C<mg_get> on C<sv> if
8319appropriate, else not. C<sv_pvn_force> and C<sv_pvn_force_nomg> are
8320implemented in terms of this function.
645c22ef
DM
8321You normally want to use the various wrapper macros instead: see
8322C<SvPV_force> and C<SvPV_force_nomg>
8d6d96c1
HS
8323
8324=cut
8325*/
8326
8327char *
12964ddd 8328Perl_sv_pvn_force_flags(pTHX_ SV *const sv, STRLEN *const lp, const I32 flags)
8d6d96c1 8329{
97aff369 8330 dVAR;
7918f24d
NC
8331
8332 PERL_ARGS_ASSERT_SV_PVN_FORCE_FLAGS;
8333
6fc92669 8334 if (SvTHINKFIRST(sv) && !SvROK(sv))
765f542d 8335 sv_force_normal_flags(sv, 0);
1c846c1f 8336
a0d0e21e 8337 if (SvPOK(sv)) {
13c5b33c
NC
8338 if (lp)
8339 *lp = SvCUR(sv);
a0d0e21e
LW
8340 }
8341 else {
a3b680e6 8342 char *s;
13c5b33c
NC
8343 STRLEN len;
8344
4d84ee25 8345 if (SvREADONLY(sv) && !(flags & SV_MUTABLE_RETURN)) {
b64e5050 8346 const char * const ref = sv_reftype(sv,0);
4d84ee25
NC
8347 if (PL_op)
8348 Perl_croak(aTHX_ "Can't coerce readonly %s to string in %s",
b64e5050 8349 ref, OP_NAME(PL_op));
4d84ee25 8350 else
b64e5050 8351 Perl_croak(aTHX_ "Can't coerce readonly %s to string", ref);
4d84ee25 8352 }
1f257c95
NC
8353 if ((SvTYPE(sv) > SVt_PVLV && SvTYPE(sv) != SVt_PVFM)
8354 || isGV_with_GP(sv))
cea2e8a9 8355 Perl_croak(aTHX_ "Can't coerce %s to string in %s", sv_reftype(sv,0),
53e06cf0 8356 OP_NAME(PL_op));
b64e5050 8357 s = sv_2pv_flags(sv, &len, flags);
13c5b33c
NC
8358 if (lp)
8359 *lp = len;
8360
3f7c398e 8361 if (s != SvPVX_const(sv)) { /* Almost, but not quite, sv_setpvn() */
a0d0e21e
LW
8362 if (SvROK(sv))
8363 sv_unref(sv);
862a34c6 8364 SvUPGRADE(sv, SVt_PV); /* Never FALSE */
a0d0e21e 8365 SvGROW(sv, len + 1);
706aa1c9 8366 Move(s,SvPVX(sv),len,char);
a0d0e21e 8367 SvCUR_set(sv, len);
97a130b8 8368 SvPVX(sv)[len] = '\0';
a0d0e21e
LW
8369 }
8370 if (!SvPOK(sv)) {
8371 SvPOK_on(sv); /* validate pointer */
8372 SvTAINT(sv);
1d7c1841 8373 DEBUG_c(PerlIO_printf(Perl_debug_log, "0x%"UVxf" 2pv(%s)\n",
3f7c398e 8374 PTR2UV(sv),SvPVX_const(sv)));
a0d0e21e
LW
8375 }
8376 }
4d84ee25 8377 return SvPVX_mutable(sv);
a0d0e21e
LW
8378}
8379
645c22ef 8380/*
645c22ef
DM
8381=for apidoc sv_pvbyten_force
8382
0feed65a 8383The backend for the C<SvPVbytex_force> macro. Always use the macro instead.
645c22ef
DM
8384
8385=cut
8386*/
8387
7340a771 8388char *
12964ddd 8389Perl_sv_pvbyten_force(pTHX_ SV *const sv, STRLEN *const lp)
7340a771 8390{
7918f24d
NC
8391 PERL_ARGS_ASSERT_SV_PVBYTEN_FORCE;
8392
46ec2f14 8393 sv_pvn_force(sv,lp);
ffebcc3e 8394 sv_utf8_downgrade(sv,0);
46ec2f14
TS
8395 *lp = SvCUR(sv);
8396 return SvPVX(sv);
7340a771
GS
8397}
8398
645c22ef 8399/*
c461cf8f
JH
8400=for apidoc sv_pvutf8n_force
8401
0feed65a 8402The backend for the C<SvPVutf8x_force> macro. Always use the macro instead.
c461cf8f
JH
8403
8404=cut
8405*/
8406
7340a771 8407char *
12964ddd 8408Perl_sv_pvutf8n_force(pTHX_ SV *const sv, STRLEN *const lp)
7340a771 8409{
7918f24d
NC
8410 PERL_ARGS_ASSERT_SV_PVUTF8N_FORCE;
8411
46ec2f14 8412 sv_pvn_force(sv,lp);
560a288e 8413 sv_utf8_upgrade(sv);
46ec2f14
TS
8414 *lp = SvCUR(sv);
8415 return SvPVX(sv);
7340a771
GS
8416}
8417
c461cf8f
JH
8418/*
8419=for apidoc sv_reftype
8420
8421Returns a string describing what the SV is a reference to.
8422
8423=cut
8424*/
8425
2b388283 8426const char *
12964ddd 8427Perl_sv_reftype(pTHX_ const SV *const sv, const int ob)
a0d0e21e 8428{
7918f24d
NC
8429 PERL_ARGS_ASSERT_SV_REFTYPE;
8430
07409e01
NC
8431 /* The fact that I don't need to downcast to char * everywhere, only in ?:
8432 inside return suggests a const propagation bug in g++. */
c86bf373 8433 if (ob && SvOBJECT(sv)) {
1b6737cc 8434 char * const name = HvNAME_get(SvSTASH(sv));
07409e01 8435 return name ? name : (char *) "__ANON__";
c86bf373 8436 }
a0d0e21e
LW
8437 else {
8438 switch (SvTYPE(sv)) {
8439 case SVt_NULL:
8440 case SVt_IV:
8441 case SVt_NV:
a0d0e21e
LW
8442 case SVt_PV:
8443 case SVt_PVIV:
8444 case SVt_PVNV:
8445 case SVt_PVMG:
1cb0ed9b 8446 if (SvVOK(sv))
439cb1c4 8447 return "VSTRING";
a0d0e21e
LW
8448 if (SvROK(sv))
8449 return "REF";
8450 else
8451 return "SCALAR";
1cb0ed9b 8452
07409e01 8453 case SVt_PVLV: return (char *) (SvROK(sv) ? "REF"
be65207d
DM
8454 /* tied lvalues should appear to be
8455 * scalars for backwards compatitbility */
8456 : (LvTYPE(sv) == 't' || LvTYPE(sv) == 'T')
07409e01 8457 ? "SCALAR" : "LVALUE");
a0d0e21e
LW
8458 case SVt_PVAV: return "ARRAY";
8459 case SVt_PVHV: return "HASH";
8460 case SVt_PVCV: return "CODE";
6e592b3a
BM
8461 case SVt_PVGV: return (char *) (isGV_with_GP(sv)
8462 ? "GLOB" : "SCALAR");
1d2dff63 8463 case SVt_PVFM: return "FORMAT";
27f9d8f3 8464 case SVt_PVIO: return "IO";
cecf5685 8465 case SVt_BIND: return "BIND";
b7c9370f 8466 case SVt_REGEXP: return "REGEXP";
a0d0e21e
LW
8467 default: return "UNKNOWN";
8468 }
8469 }
8470}
8471
954c1994
GS
8472/*
8473=for apidoc sv_isobject
8474
8475Returns a boolean indicating whether the SV is an RV pointing to a blessed
8476object. If the SV is not an RV, or if the object is not blessed, then this
8477will return false.
8478
8479=cut
8480*/
8481
463ee0b2 8482int
864dbfa3 8483Perl_sv_isobject(pTHX_ SV *sv)
85e6fe83 8484{
68dc0745 8485 if (!sv)
8486 return 0;
5b295bef 8487 SvGETMAGIC(sv);
85e6fe83
LW
8488 if (!SvROK(sv))
8489 return 0;
daba3364 8490 sv = SvRV(sv);
85e6fe83
LW
8491 if (!SvOBJECT(sv))
8492 return 0;
8493 return 1;
8494}
8495
954c1994
GS
8496/*
8497=for apidoc sv_isa
8498
8499Returns a boolean indicating whether the SV is blessed into the specified
8500class. This does not check for subtypes; use C<sv_derived_from> to verify
8501an inheritance relationship.
8502
8503=cut
8504*/
8505
85e6fe83 8506int
12964ddd 8507Perl_sv_isa(pTHX_ SV *sv, const char *const name)
463ee0b2 8508{
bfcb3514 8509 const char *hvname;
7918f24d
NC
8510
8511 PERL_ARGS_ASSERT_SV_ISA;
8512
68dc0745 8513 if (!sv)
8514 return 0;
5b295bef 8515 SvGETMAGIC(sv);
ed6116ce 8516 if (!SvROK(sv))
463ee0b2 8517 return 0;
daba3364 8518 sv = SvRV(sv);
ed6116ce 8519 if (!SvOBJECT(sv))
463ee0b2 8520 return 0;
bfcb3514
NC
8521 hvname = HvNAME_get(SvSTASH(sv));
8522 if (!hvname)
e27ad1f2 8523 return 0;
463ee0b2 8524
bfcb3514 8525 return strEQ(hvname, name);
463ee0b2
LW
8526}
8527
954c1994
GS
8528/*
8529=for apidoc newSVrv
8530
8531Creates a new SV for the RV, C<rv>, to point to. If C<rv> is not an RV then
8532it will be upgraded to one. If C<classname> is non-null then the new SV will
8533be blessed in the specified package. The new SV is returned and its
8534reference count is 1.
8535
8536=cut
8537*/
8538
463ee0b2 8539SV*
12964ddd 8540Perl_newSVrv(pTHX_ SV *const rv, const char *const classname)
463ee0b2 8541{
97aff369 8542 dVAR;
463ee0b2
LW
8543 SV *sv;
8544
7918f24d
NC
8545 PERL_ARGS_ASSERT_NEWSVRV;
8546
4561caa4 8547 new_SV(sv);
51cf62d8 8548
765f542d 8549 SV_CHECK_THINKFIRST_COW_DROP(rv);
52944de8 8550 (void)SvAMAGIC_off(rv);
51cf62d8 8551
0199fce9 8552 if (SvTYPE(rv) >= SVt_PVMG) {
a3b680e6 8553 const U32 refcnt = SvREFCNT(rv);
0199fce9
JD
8554 SvREFCNT(rv) = 0;
8555 sv_clear(rv);
8556 SvFLAGS(rv) = 0;
8557 SvREFCNT(rv) = refcnt;
0199fce9 8558
4df7f6af 8559 sv_upgrade(rv, SVt_IV);
dc5494d2
NC
8560 } else if (SvROK(rv)) {
8561 SvREFCNT_dec(SvRV(rv));
43230e26
NC
8562 } else {
8563 prepare_SV_for_RV(rv);
0199fce9 8564 }
51cf62d8 8565
0c34ef67 8566 SvOK_off(rv);
b162af07 8567 SvRV_set(rv, sv);
ed6116ce 8568 SvROK_on(rv);
463ee0b2 8569
a0d0e21e 8570 if (classname) {
da51bb9b 8571 HV* const stash = gv_stashpv(classname, GV_ADD);
a0d0e21e
LW
8572 (void)sv_bless(rv, stash);
8573 }
8574 return sv;
8575}
8576
954c1994
GS
8577/*
8578=for apidoc sv_setref_pv
8579
8580Copies a pointer into a new SV, optionally blessing the SV. The C<rv>
8581argument will be upgraded to an RV. That RV will be modified to point to
8582the new SV. If the C<pv> argument is NULL then C<PL_sv_undef> will be placed
8583into the SV. The C<classname> argument indicates the package for the
bd61b366 8584blessing. Set C<classname> to C<NULL> to avoid the blessing. The new SV
d34c2299 8585will have a reference count of 1, and the RV will be returned.
954c1994
GS
8586
8587Do not use with other Perl types such as HV, AV, SV, CV, because those
8588objects will become corrupted by the pointer copy process.
8589
8590Note that C<sv_setref_pvn> copies the string while this copies the pointer.
8591
8592=cut
8593*/
8594
a0d0e21e 8595SV*
12964ddd 8596Perl_sv_setref_pv(pTHX_ SV *const rv, const char *const classname, void *const pv)
a0d0e21e 8597{
97aff369 8598 dVAR;
7918f24d
NC
8599
8600 PERL_ARGS_ASSERT_SV_SETREF_PV;
8601
189b2af5 8602 if (!pv) {
3280af22 8603 sv_setsv(rv, &PL_sv_undef);
189b2af5
GS
8604 SvSETMAGIC(rv);
8605 }
a0d0e21e 8606 else
56431972 8607 sv_setiv(newSVrv(rv,classname), PTR2IV(pv));
a0d0e21e
LW
8608 return rv;
8609}
8610
954c1994
GS
8611/*
8612=for apidoc sv_setref_iv
8613
8614Copies an integer into a new SV, optionally blessing the SV. The C<rv>
8615argument will be upgraded to an RV. That RV will be modified to point to
8616the new SV. The C<classname> argument indicates the package for the
bd61b366 8617blessing. Set C<classname> to C<NULL> to avoid the blessing. The new SV
d34c2299 8618will have a reference count of 1, and the RV will be returned.
954c1994
GS
8619
8620=cut
8621*/
8622
a0d0e21e 8623SV*
12964ddd 8624Perl_sv_setref_iv(pTHX_ SV *const rv, const char *const classname, const IV iv)
a0d0e21e 8625{
7918f24d
NC
8626 PERL_ARGS_ASSERT_SV_SETREF_IV;
8627
a0d0e21e
LW
8628 sv_setiv(newSVrv(rv,classname), iv);
8629 return rv;
8630}
8631
954c1994 8632/*
e1c57cef
JH
8633=for apidoc sv_setref_uv
8634
8635Copies an unsigned integer into a new SV, optionally blessing the SV. The C<rv>
8636argument will be upgraded to an RV. That RV will be modified to point to
8637the new SV. The C<classname> argument indicates the package for the
bd61b366 8638blessing. Set C<classname> to C<NULL> to avoid the blessing. The new SV
d34c2299 8639will have a reference count of 1, and the RV will be returned.
e1c57cef
JH
8640
8641=cut
8642*/
8643
8644SV*
12964ddd 8645Perl_sv_setref_uv(pTHX_ SV *const rv, const char *const classname, const UV uv)
e1c57cef 8646{
7918f24d
NC
8647 PERL_ARGS_ASSERT_SV_SETREF_UV;
8648
e1c57cef
JH
8649 sv_setuv(newSVrv(rv,classname), uv);
8650 return rv;
8651}
8652
8653/*
954c1994
GS
8654=for apidoc sv_setref_nv
8655
8656Copies a double into a new SV, optionally blessing the SV. The C<rv>
8657argument will be upgraded to an RV. That RV will be modified to point to
8658the new SV. The C<classname> argument indicates the package for the
bd61b366 8659blessing. Set C<classname> to C<NULL> to avoid the blessing. The new SV
d34c2299 8660will have a reference count of 1, and the RV will be returned.
954c1994
GS
8661
8662=cut
8663*/
8664
a0d0e21e 8665SV*
12964ddd 8666Perl_sv_setref_nv(pTHX_ SV *const rv, const char *const classname, const NV nv)
a0d0e21e 8667{
7918f24d
NC
8668 PERL_ARGS_ASSERT_SV_SETREF_NV;
8669
a0d0e21e
LW
8670 sv_setnv(newSVrv(rv,classname), nv);
8671 return rv;
8672}
463ee0b2 8673
954c1994
GS
8674/*
8675=for apidoc sv_setref_pvn
8676
8677Copies a string into a new SV, optionally blessing the SV. The length of the
8678string must be specified with C<n>. The C<rv> argument will be upgraded to
8679an RV. That RV will be modified to point to the new SV. The C<classname>
8680argument indicates the package for the blessing. Set C<classname> to
bd61b366 8681C<NULL> to avoid the blessing. The new SV will have a reference count
d34c2299 8682of 1, and the RV will be returned.
954c1994
GS
8683
8684Note that C<sv_setref_pv> copies the pointer while this copies the string.
8685
8686=cut
8687*/
8688
a0d0e21e 8689SV*
12964ddd
SS
8690Perl_sv_setref_pvn(pTHX_ SV *const rv, const char *const classname,
8691 const char *const pv, const STRLEN n)
a0d0e21e 8692{
7918f24d
NC
8693 PERL_ARGS_ASSERT_SV_SETREF_PVN;
8694
a0d0e21e 8695 sv_setpvn(newSVrv(rv,classname), pv, n);
463ee0b2
LW
8696 return rv;
8697}
8698
954c1994
GS
8699/*
8700=for apidoc sv_bless
8701
8702Blesses an SV into a specified package. The SV must be an RV. The package
8703must be designated by its stash (see C<gv_stashpv()>). The reference count
8704of the SV is unaffected.
8705
8706=cut
8707*/
8708
a0d0e21e 8709SV*
12964ddd 8710Perl_sv_bless(pTHX_ SV *const sv, HV *const stash)
a0d0e21e 8711{
97aff369 8712 dVAR;
76e3520e 8713 SV *tmpRef;
7918f24d
NC
8714
8715 PERL_ARGS_ASSERT_SV_BLESS;
8716
a0d0e21e 8717 if (!SvROK(sv))
cea2e8a9 8718 Perl_croak(aTHX_ "Can't bless non-reference value");
76e3520e
GS
8719 tmpRef = SvRV(sv);
8720 if (SvFLAGS(tmpRef) & (SVs_OBJECT|SVf_READONLY)) {
e0744413
NC
8721 if (SvIsCOW(tmpRef))
8722 sv_force_normal_flags(tmpRef, 0);
76e3520e 8723 if (SvREADONLY(tmpRef))
f1f66076 8724 Perl_croak(aTHX_ "%s", PL_no_modify);
76e3520e
GS
8725 if (SvOBJECT(tmpRef)) {
8726 if (SvTYPE(tmpRef) != SVt_PVIO)
3280af22 8727 --PL_sv_objcount;
76e3520e 8728 SvREFCNT_dec(SvSTASH(tmpRef));
2e3febc6 8729 }
a0d0e21e 8730 }
76e3520e
GS
8731 SvOBJECT_on(tmpRef);
8732 if (SvTYPE(tmpRef) != SVt_PVIO)
3280af22 8733 ++PL_sv_objcount;
862a34c6 8734 SvUPGRADE(tmpRef, SVt_PVMG);
85fbaab2 8735 SvSTASH_set(tmpRef, MUTABLE_HV(SvREFCNT_inc_simple(stash)));
a0d0e21e 8736
2e3febc6
CS
8737 if (Gv_AMG(stash))
8738 SvAMAGIC_on(sv);
8739 else
52944de8 8740 (void)SvAMAGIC_off(sv);
a0d0e21e 8741
1edbfb88
AB
8742 if(SvSMAGICAL(tmpRef))
8743 if(mg_find(tmpRef, PERL_MAGIC_ext) || mg_find(tmpRef, PERL_MAGIC_uvar))
8744 mg_set(tmpRef);
8745
8746
ecdeb87c 8747
a0d0e21e
LW
8748 return sv;
8749}
8750
645c22ef 8751/* Downgrades a PVGV to a PVMG.
645c22ef
DM
8752 */
8753
76e3520e 8754STATIC void
89e38212 8755S_sv_unglob(pTHX_ SV *const sv)
a0d0e21e 8756{
97aff369 8757 dVAR;
850fabdf 8758 void *xpvmg;
dd69841b 8759 HV *stash;
b37c2d43 8760 SV * const temp = sv_newmortal();
850fabdf 8761
7918f24d
NC
8762 PERL_ARGS_ASSERT_SV_UNGLOB;
8763
a0d0e21e
LW
8764 assert(SvTYPE(sv) == SVt_PVGV);
8765 SvFAKE_off(sv);
159b6efe 8766 gv_efullname3(temp, MUTABLE_GV(sv), "*");
180488f8 8767
f7877b28 8768 if (GvGP(sv)) {
159b6efe
NC
8769 if(GvCVu((const GV *)sv) && (stash = GvSTASH(MUTABLE_GV(sv)))
8770 && HvNAME_get(stash))
dd69841b 8771 mro_method_changed_in(stash);
159b6efe 8772 gp_free(MUTABLE_GV(sv));
f7877b28 8773 }
e826b3c7 8774 if (GvSTASH(sv)) {
daba3364 8775 sv_del_backref(MUTABLE_SV(GvSTASH(sv)), sv);
5c284bb0 8776 GvSTASH(sv) = NULL;
e826b3c7 8777 }
a5f75d66 8778 GvMULTI_off(sv);
acda4c6a
NC
8779 if (GvNAME_HEK(sv)) {
8780 unshare_hek(GvNAME_HEK(sv));
8781 }
2e5b91de 8782 isGV_with_GP_off(sv);
850fabdf
GS
8783
8784 /* need to keep SvANY(sv) in the right arena */
8785 xpvmg = new_XPVMG();
8786 StructCopy(SvANY(sv), xpvmg, XPVMG);
8787 del_XPVGV(SvANY(sv));
8788 SvANY(sv) = xpvmg;
8789
a0d0e21e
LW
8790 SvFLAGS(sv) &= ~SVTYPEMASK;
8791 SvFLAGS(sv) |= SVt_PVMG;
180488f8
NC
8792
8793 /* Intentionally not calling any local SET magic, as this isn't so much a
8794 set operation as merely an internal storage change. */
8795 sv_setsv_flags(sv, temp, 0);
a0d0e21e
LW
8796}
8797
954c1994 8798/*
840a7b70 8799=for apidoc sv_unref_flags
954c1994
GS
8800
8801Unsets the RV status of the SV, and decrements the reference count of
8802whatever was being referenced by the RV. This can almost be thought of
840a7b70
IZ
8803as a reversal of C<newSVrv>. The C<cflags> argument can contain
8804C<SV_IMMEDIATE_UNREF> to force the reference count to be decremented
8805(otherwise the decrementing is conditional on the reference count being
8806different from one or the reference being a readonly SV).
7889fe52 8807See C<SvROK_off>.
954c1994
GS
8808
8809=cut
8810*/
8811
ed6116ce 8812void
89e38212 8813Perl_sv_unref_flags(pTHX_ SV *const ref, const U32 flags)
ed6116ce 8814{
b64e5050 8815 SV* const target = SvRV(ref);
810b8aa5 8816
7918f24d
NC
8817 PERL_ARGS_ASSERT_SV_UNREF_FLAGS;
8818
e15faf7d
NC
8819 if (SvWEAKREF(ref)) {
8820 sv_del_backref(target, ref);
8821 SvWEAKREF_off(ref);
8822 SvRV_set(ref, NULL);
810b8aa5
GS
8823 return;
8824 }
e15faf7d
NC
8825 SvRV_set(ref, NULL);
8826 SvROK_off(ref);
8827 /* You can't have a || SvREADONLY(target) here, as $a = $$a, where $a was
04ca4930 8828 assigned to as BEGIN {$a = \"Foo"} will fail. */
e15faf7d
NC
8829 if (SvREFCNT(target) != 1 || (flags & SV_IMMEDIATE_UNREF))
8830 SvREFCNT_dec(target);
840a7b70 8831 else /* XXX Hack, but hard to make $a=$a->[1] work otherwise */
e15faf7d 8832 sv_2mortal(target); /* Schedule for freeing later */
ed6116ce 8833}
8990e307 8834
840a7b70 8835/*
645c22ef
DM
8836=for apidoc sv_untaint
8837
8838Untaint an SV. Use C<SvTAINTED_off> instead.
8839=cut
8840*/
8841
bbce6d69 8842void
89e38212 8843Perl_sv_untaint(pTHX_ SV *const sv)
bbce6d69 8844{
7918f24d
NC
8845 PERL_ARGS_ASSERT_SV_UNTAINT;
8846
13f57bf8 8847 if (SvTYPE(sv) >= SVt_PVMG && SvMAGIC(sv)) {
b64e5050 8848 MAGIC * const mg = mg_find(sv, PERL_MAGIC_taint);
36477c24 8849 if (mg)
565764a8 8850 mg->mg_len &= ~1;
36477c24 8851 }
bbce6d69 8852}
8853
645c22ef
DM
8854/*
8855=for apidoc sv_tainted
8856
8857Test an SV for taintedness. Use C<SvTAINTED> instead.
8858=cut
8859*/
8860
bbce6d69 8861bool
89e38212 8862Perl_sv_tainted(pTHX_ SV *const sv)
bbce6d69 8863{
7918f24d
NC
8864 PERL_ARGS_ASSERT_SV_TAINTED;
8865
13f57bf8 8866 if (SvTYPE(sv) >= SVt_PVMG && SvMAGIC(sv)) {
823a54a3 8867 const MAGIC * const mg = mg_find(sv, PERL_MAGIC_taint);
2ddb8a4f 8868 if (mg && (mg->mg_len & 1) )
36477c24 8869 return TRUE;
8870 }
8871 return FALSE;
bbce6d69 8872}
8873
09540bc3
JH
8874/*
8875=for apidoc sv_setpviv
8876
8877Copies an integer into the given SV, also updating its string value.
8878Does not handle 'set' magic. See C<sv_setpviv_mg>.
8879
8880=cut
8881*/
8882
8883void
89e38212 8884Perl_sv_setpviv(pTHX_ SV *const sv, const IV iv)
09540bc3
JH
8885{
8886 char buf[TYPE_CHARS(UV)];
8887 char *ebuf;
b64e5050 8888 char * const ptr = uiv_2buf(buf, iv, 0, 0, &ebuf);
09540bc3 8889
7918f24d
NC
8890 PERL_ARGS_ASSERT_SV_SETPVIV;
8891
09540bc3
JH
8892 sv_setpvn(sv, ptr, ebuf - ptr);
8893}
8894
8895/*
8896=for apidoc sv_setpviv_mg
8897
8898Like C<sv_setpviv>, but also handles 'set' magic.
8899
8900=cut
8901*/
8902
8903void
89e38212 8904Perl_sv_setpviv_mg(pTHX_ SV *const sv, const IV iv)
09540bc3 8905{
7918f24d
NC
8906 PERL_ARGS_ASSERT_SV_SETPVIV_MG;
8907
df7eb254 8908 sv_setpviv(sv, iv);
09540bc3
JH
8909 SvSETMAGIC(sv);
8910}
8911
cea2e8a9 8912#if defined(PERL_IMPLICIT_CONTEXT)
645c22ef
DM
8913
8914/* pTHX_ magic can't cope with varargs, so this is a no-context
8915 * version of the main function, (which may itself be aliased to us).
8916 * Don't access this version directly.
8917 */
8918
cea2e8a9 8919void
89e38212 8920Perl_sv_setpvf_nocontext(SV *const sv, const char *const pat, ...)
cea2e8a9
GS
8921{
8922 dTHX;
8923 va_list args;
7918f24d
NC
8924
8925 PERL_ARGS_ASSERT_SV_SETPVF_NOCONTEXT;
8926
cea2e8a9 8927 va_start(args, pat);
c5be433b 8928 sv_vsetpvf(sv, pat, &args);
cea2e8a9
GS
8929 va_end(args);
8930}
8931
645c22ef
DM
8932/* pTHX_ magic can't cope with varargs, so this is a no-context
8933 * version of the main function, (which may itself be aliased to us).
8934 * Don't access this version directly.
8935 */
cea2e8a9
GS
8936
8937void
89e38212 8938Perl_sv_setpvf_mg_nocontext(SV *const sv, const char *const pat, ...)
cea2e8a9
GS
8939{
8940 dTHX;
8941 va_list args;
7918f24d
NC
8942
8943 PERL_ARGS_ASSERT_SV_SETPVF_MG_NOCONTEXT;
8944
cea2e8a9 8945 va_start(args, pat);
c5be433b 8946 sv_vsetpvf_mg(sv, pat, &args);
cea2e8a9 8947 va_end(args);
cea2e8a9
GS
8948}
8949#endif
8950
954c1994
GS
8951/*
8952=for apidoc sv_setpvf
8953
bffc3d17
SH
8954Works like C<sv_catpvf> but copies the text into the SV instead of
8955appending it. Does not handle 'set' magic. See C<sv_setpvf_mg>.
954c1994
GS
8956
8957=cut
8958*/
8959
46fc3d4c 8960void
89e38212 8961Perl_sv_setpvf(pTHX_ SV *const sv, const char *const pat, ...)
46fc3d4c 8962{
8963 va_list args;
7918f24d
NC
8964
8965 PERL_ARGS_ASSERT_SV_SETPVF;
8966
46fc3d4c 8967 va_start(args, pat);
c5be433b 8968 sv_vsetpvf(sv, pat, &args);
46fc3d4c 8969 va_end(args);
8970}
8971
bffc3d17
SH
8972/*
8973=for apidoc sv_vsetpvf
8974
8975Works like C<sv_vcatpvf> but copies the text into the SV instead of
8976appending it. Does not handle 'set' magic. See C<sv_vsetpvf_mg>.
8977
8978Usually used via its frontend C<sv_setpvf>.
8979
8980=cut
8981*/
645c22ef 8982
c5be433b 8983void
89e38212 8984Perl_sv_vsetpvf(pTHX_ SV *const sv, const char *const pat, va_list *const args)
c5be433b 8985{
7918f24d
NC
8986 PERL_ARGS_ASSERT_SV_VSETPVF;
8987
4608196e 8988 sv_vsetpvfn(sv, pat, strlen(pat), args, NULL, 0, NULL);
c5be433b 8989}
ef50df4b 8990
954c1994
GS
8991/*
8992=for apidoc sv_setpvf_mg
8993
8994Like C<sv_setpvf>, but also handles 'set' magic.
8995
8996=cut
8997*/
8998
ef50df4b 8999void
89e38212 9000Perl_sv_setpvf_mg(pTHX_ SV *const sv, const char *const pat, ...)
ef50df4b
GS
9001{
9002 va_list args;
7918f24d
NC
9003
9004 PERL_ARGS_ASSERT_SV_SETPVF_MG;
9005
ef50df4b 9006 va_start(args, pat);
c5be433b 9007 sv_vsetpvf_mg(sv, pat, &args);
ef50df4b 9008 va_end(args);
c5be433b
GS
9009}
9010
bffc3d17
SH
9011/*
9012=for apidoc sv_vsetpvf_mg
9013
9014Like C<sv_vsetpvf>, but also handles 'set' magic.
9015
9016Usually used via its frontend C<sv_setpvf_mg>.
9017
9018=cut
9019*/
645c22ef 9020
c5be433b 9021void
89e38212 9022Perl_sv_vsetpvf_mg(pTHX_ SV *const sv, const char *const pat, va_list *const args)
c5be433b 9023{
7918f24d
NC
9024 PERL_ARGS_ASSERT_SV_VSETPVF_MG;
9025
4608196e 9026 sv_vsetpvfn(sv, pat, strlen(pat), args, NULL, 0, NULL);
ef50df4b
GS
9027 SvSETMAGIC(sv);
9028}
9029
cea2e8a9 9030#if defined(PERL_IMPLICIT_CONTEXT)
645c22ef
DM
9031
9032/* pTHX_ magic can't cope with varargs, so this is a no-context
9033 * version of the main function, (which may itself be aliased to us).
9034 * Don't access this version directly.
9035 */
9036
cea2e8a9 9037void
89e38212 9038Perl_sv_catpvf_nocontext(SV *const sv, const char *const pat, ...)
cea2e8a9
GS
9039{
9040 dTHX;
9041 va_list args;
7918f24d
NC
9042
9043 PERL_ARGS_ASSERT_SV_CATPVF_NOCONTEXT;
9044
cea2e8a9 9045 va_start(args, pat);
c5be433b 9046 sv_vcatpvf(sv, pat, &args);
cea2e8a9
GS
9047 va_end(args);
9048}
9049
645c22ef
DM
9050/* pTHX_ magic can't cope with varargs, so this is a no-context
9051 * version of the main function, (which may itself be aliased to us).
9052 * Don't access this version directly.
9053 */
9054
cea2e8a9 9055void
89e38212 9056Perl_sv_catpvf_mg_nocontext(SV *const sv, const char *const pat, ...)
cea2e8a9
GS
9057{
9058 dTHX;
9059 va_list args;
7918f24d
NC
9060
9061 PERL_ARGS_ASSERT_SV_CATPVF_MG_NOCONTEXT;
9062
cea2e8a9 9063 va_start(args, pat);
c5be433b 9064 sv_vcatpvf_mg(sv, pat, &args);
cea2e8a9 9065 va_end(args);
cea2e8a9
GS
9066}
9067#endif
9068
954c1994
GS
9069/*
9070=for apidoc sv_catpvf
9071
d5ce4a7c
GA
9072Processes its arguments like C<sprintf> and appends the formatted
9073output to an SV. If the appended data contains "wide" characters
9074(including, but not limited to, SVs with a UTF-8 PV formatted with %s,
9075and characters >255 formatted with %c), the original SV might get
bffc3d17 9076upgraded to UTF-8. Handles 'get' magic, but not 'set' magic. See
cdd94ca7
NC
9077C<sv_catpvf_mg>. If the original SV was UTF-8, the pattern should be
9078valid UTF-8; if the original SV was bytes, the pattern should be too.
954c1994 9079
d5ce4a7c 9080=cut */
954c1994 9081
46fc3d4c 9082void
66ceb532 9083Perl_sv_catpvf(pTHX_ SV *const sv, const char *const pat, ...)
46fc3d4c 9084{
9085 va_list args;
7918f24d
NC
9086
9087 PERL_ARGS_ASSERT_SV_CATPVF;
9088
46fc3d4c 9089 va_start(args, pat);
c5be433b 9090 sv_vcatpvf(sv, pat, &args);
46fc3d4c 9091 va_end(args);
9092}
9093
bffc3d17
SH
9094/*
9095=for apidoc sv_vcatpvf
9096
9097Processes its arguments like C<vsprintf> and appends the formatted output
9098to an SV. Does not handle 'set' magic. See C<sv_vcatpvf_mg>.
9099
9100Usually used via its frontend C<sv_catpvf>.
9101
9102=cut
9103*/
645c22ef 9104
ef50df4b 9105void
66ceb532 9106Perl_sv_vcatpvf(pTHX_ SV *const sv, const char *const pat, va_list *const args)
c5be433b 9107{
7918f24d
NC
9108 PERL_ARGS_ASSERT_SV_VCATPVF;
9109
4608196e 9110 sv_vcatpvfn(sv, pat, strlen(pat), args, NULL, 0, NULL);
c5be433b
GS
9111}
9112
954c1994
GS
9113/*
9114=for apidoc sv_catpvf_mg
9115
9116Like C<sv_catpvf>, but also handles 'set' magic.
9117
9118=cut
9119*/
9120
c5be433b 9121void
66ceb532 9122Perl_sv_catpvf_mg(pTHX_ SV *const sv, const char *const pat, ...)
ef50df4b
GS
9123{
9124 va_list args;
7918f24d
NC
9125
9126 PERL_ARGS_ASSERT_SV_CATPVF_MG;
9127
ef50df4b 9128 va_start(args, pat);
c5be433b 9129 sv_vcatpvf_mg(sv, pat, &args);
ef50df4b 9130 va_end(args);
c5be433b
GS
9131}
9132
bffc3d17
SH
9133/*
9134=for apidoc sv_vcatpvf_mg
9135
9136Like C<sv_vcatpvf>, but also handles 'set' magic.
9137
9138Usually used via its frontend C<sv_catpvf_mg>.
9139
9140=cut
9141*/
645c22ef 9142
c5be433b 9143void
66ceb532 9144Perl_sv_vcatpvf_mg(pTHX_ SV *const sv, const char *const pat, va_list *const args)
c5be433b 9145{
7918f24d
NC
9146 PERL_ARGS_ASSERT_SV_VCATPVF_MG;
9147
4608196e 9148 sv_vcatpvfn(sv, pat, strlen(pat), args, NULL, 0, NULL);
ef50df4b
GS
9149 SvSETMAGIC(sv);
9150}
9151
954c1994
GS
9152/*
9153=for apidoc sv_vsetpvfn
9154
bffc3d17 9155Works like C<sv_vcatpvfn> but copies the text into the SV instead of
954c1994
GS
9156appending it.
9157
bffc3d17 9158Usually used via one of its frontends C<sv_vsetpvf> and C<sv_vsetpvf_mg>.
645c22ef 9159
954c1994
GS
9160=cut
9161*/
9162
46fc3d4c 9163void
66ceb532
SS
9164Perl_sv_vsetpvfn(pTHX_ SV *const sv, const char *const pat, const STRLEN patlen,
9165 va_list *const args, SV **const svargs, const I32 svmax, bool *const maybe_tainted)
46fc3d4c 9166{
7918f24d
NC
9167 PERL_ARGS_ASSERT_SV_VSETPVFN;
9168
76f68e9b 9169 sv_setpvs(sv, "");
7d5ea4e7 9170 sv_vcatpvfn(sv, pat, patlen, args, svargs, svmax, maybe_tainted);
46fc3d4c 9171}
9172
7baa4690
HS
9173
9174/*
9175 * Warn of missing argument to sprintf, and then return a defined value
9176 * to avoid inappropriate "use of uninit" warnings [perl #71000].
9177 */
9178#define WARN_MISSING WARN_UNINITIALIZED /* Not sure we want a new category */
9179STATIC SV*
81ae3cde 9180S_vcatpvfn_missing_argument(pTHX) {
7baa4690
HS
9181 if (ckWARN(WARN_MISSING)) {
9182 Perl_warner(aTHX_ packWARN(WARN_MISSING), "Missing argument in %s",
9183 PL_op ? OP_DESC(PL_op) : "sv_vcatpvfn()");
9184 }
9185 return &PL_sv_no;
9186}
9187
9188
2d00ba3b 9189STATIC I32
66ceb532 9190S_expect_number(pTHX_ char **const pattern)
211dfcf1 9191{
97aff369 9192 dVAR;
211dfcf1 9193 I32 var = 0;
7918f24d
NC
9194
9195 PERL_ARGS_ASSERT_EXPECT_NUMBER;
9196
211dfcf1
HS
9197 switch (**pattern) {
9198 case '1': case '2': case '3':
9199 case '4': case '5': case '6':
9200 case '7': case '8': case '9':
2fba7546
GA
9201 var = *(*pattern)++ - '0';
9202 while (isDIGIT(**pattern)) {
5f66b61c 9203 const I32 tmp = var * 10 + (*(*pattern)++ - '0');
2fba7546
GA
9204 if (tmp < var)
9205 Perl_croak(aTHX_ "Integer overflow in format string for %s", (PL_op ? OP_NAME(PL_op) : "sv_vcatpvfn"));
9206 var = tmp;
9207 }
211dfcf1
HS
9208 }
9209 return var;
9210}
211dfcf1 9211
c445ea15 9212STATIC char *
66ceb532 9213S_F0convert(NV nv, char *const endbuf, STRLEN *const len)
4151a5fe 9214{
a3b680e6 9215 const int neg = nv < 0;
4151a5fe 9216 UV uv;
4151a5fe 9217
7918f24d
NC
9218 PERL_ARGS_ASSERT_F0CONVERT;
9219
4151a5fe
IZ
9220 if (neg)
9221 nv = -nv;
9222 if (nv < UV_MAX) {
b464bac0 9223 char *p = endbuf;
4151a5fe 9224 nv += 0.5;
028f8eaa 9225 uv = (UV)nv;
4151a5fe
IZ
9226 if (uv & 1 && uv == nv)
9227 uv--; /* Round to even */
9228 do {
a3b680e6 9229 const unsigned dig = uv % 10;
4151a5fe
IZ
9230 *--p = '0' + dig;
9231 } while (uv /= 10);
9232 if (neg)
9233 *--p = '-';
9234 *len = endbuf - p;
9235 return p;
9236 }
bd61b366 9237 return NULL;
4151a5fe
IZ
9238}
9239
9240
954c1994
GS
9241/*
9242=for apidoc sv_vcatpvfn
9243
9244Processes its arguments like C<vsprintf> and appends the formatted output
9245to an SV. Uses an array of SVs if the C style variable argument list is
9246missing (NULL). When running with taint checks enabled, indicates via
9247C<maybe_tainted> if results are untrustworthy (often due to the use of
9248locales).
9249
bffc3d17 9250Usually used via one of its frontends C<sv_vcatpvf> and C<sv_vcatpvf_mg>.
645c22ef 9251
954c1994
GS
9252=cut
9253*/
9254
8896765a
RB
9255
9256#define VECTORIZE_ARGS vecsv = va_arg(*args, SV*);\
9257 vecstr = (U8*)SvPV_const(vecsv,veclen);\
9258 vec_utf8 = DO_UTF8(vecsv);
9259
1ef29b0e
RGS
9260/* XXX maybe_tainted is never assigned to, so the doc above is lying. */
9261
46fc3d4c 9262void
66ceb532
SS
9263Perl_sv_vcatpvfn(pTHX_ SV *const sv, const char *const pat, const STRLEN patlen,
9264 va_list *const args, SV **const svargs, const I32 svmax, bool *const maybe_tainted)
46fc3d4c 9265{
97aff369 9266 dVAR;
46fc3d4c 9267 char *p;
9268 char *q;
a3b680e6 9269 const char *patend;
fc36a67e 9270 STRLEN origlen;
46fc3d4c 9271 I32 svix = 0;
27da23d5 9272 static const char nullstr[] = "(null)";
a0714e2c 9273 SV *argsv = NULL;
b464bac0
AL
9274 bool has_utf8 = DO_UTF8(sv); /* has the result utf8? */
9275 const bool pat_utf8 = has_utf8; /* the pattern is in utf8? */
a0714e2c 9276 SV *nsv = NULL;
4151a5fe
IZ
9277 /* Times 4: a decimal digit takes more than 3 binary digits.
9278 * NV_DIG: mantissa takes than many decimal digits.
9279 * Plus 32: Playing safe. */
9280 char ebuf[IV_DIG * 4 + NV_DIG + 32];
9281 /* large enough for "%#.#f" --chip */
9282 /* what about long double NVs? --jhi */
db79b45b 9283
7918f24d 9284 PERL_ARGS_ASSERT_SV_VCATPVFN;
53c1dcc0
AL
9285 PERL_UNUSED_ARG(maybe_tainted);
9286
46fc3d4c 9287 /* no matter what, this is a string now */
fc36a67e 9288 (void)SvPV_force(sv, origlen);
46fc3d4c 9289
8896765a 9290 /* special-case "", "%s", and "%-p" (SVf - see below) */
46fc3d4c 9291 if (patlen == 0)
9292 return;
0dbb1585 9293 if (patlen == 2 && pat[0] == '%' && pat[1] == 's') {
2d03de9c
AL
9294 if (args) {
9295 const char * const s = va_arg(*args, char*);
9296 sv_catpv(sv, s ? s : nullstr);
9297 }
9298 else if (svix < svmax) {
9299 sv_catsv(sv, *svargs);
2d03de9c
AL
9300 }
9301 return;
0dbb1585 9302 }
8896765a
RB
9303 if (args && patlen == 3 && pat[0] == '%' &&
9304 pat[1] == '-' && pat[2] == 'p') {
daba3364 9305 argsv = MUTABLE_SV(va_arg(*args, void*));
8896765a 9306 sv_catsv(sv, argsv);
8896765a 9307 return;
46fc3d4c 9308 }
9309
1d917b39 9310#ifndef USE_LONG_DOUBLE
4151a5fe 9311 /* special-case "%.<number>[gf]" */
7af36d83 9312 if ( !args && patlen <= 5 && pat[0] == '%' && pat[1] == '.'
4151a5fe
IZ
9313 && (pat[patlen-1] == 'g' || pat[patlen-1] == 'f') ) {
9314 unsigned digits = 0;
9315 const char *pp;
9316
9317 pp = pat + 2;
9318 while (*pp >= '0' && *pp <= '9')
9319 digits = 10 * digits + (*pp++ - '0');
028f8eaa 9320 if (pp - pat == (int)patlen - 1) {
4151a5fe
IZ
9321 NV nv;
9322
7af36d83 9323 if (svix < svmax)
4151a5fe
IZ
9324 nv = SvNV(*svargs);
9325 else
9326 return;
9327 if (*pp == 'g') {
2873255c
NC
9328 /* Add check for digits != 0 because it seems that some
9329 gconverts are buggy in this case, and we don't yet have
9330 a Configure test for this. */
9331 if (digits && digits < sizeof(ebuf) - NV_DIG - 10) {
9332 /* 0, point, slack */
2e59c212 9333 Gconvert(nv, (int)digits, 0, ebuf);
4151a5fe
IZ
9334 sv_catpv(sv, ebuf);
9335 if (*ebuf) /* May return an empty string for digits==0 */
9336 return;
9337 }
9338 } else if (!digits) {
9339 STRLEN l;
9340
9341 if ((p = F0convert(nv, ebuf + sizeof ebuf, &l))) {
9342 sv_catpvn(sv, p, l);
9343 return;
9344 }
9345 }
9346 }
9347 }
1d917b39 9348#endif /* !USE_LONG_DOUBLE */
4151a5fe 9349
2cf2cfc6 9350 if (!args && svix < svmax && DO_UTF8(*svargs))
205f51d8 9351 has_utf8 = TRUE;
2cf2cfc6 9352
46fc3d4c 9353 patend = (char*)pat + patlen;
9354 for (p = (char*)pat; p < patend; p = q) {
9355 bool alt = FALSE;
9356 bool left = FALSE;
b22c7a20 9357 bool vectorize = FALSE;
211dfcf1 9358 bool vectorarg = FALSE;
2cf2cfc6 9359 bool vec_utf8 = FALSE;
46fc3d4c 9360 char fill = ' ';
9361 char plus = 0;
9362 char intsize = 0;
9363 STRLEN width = 0;
fc36a67e 9364 STRLEN zeros = 0;
46fc3d4c 9365 bool has_precis = FALSE;
9366 STRLEN precis = 0;
c445ea15 9367 const I32 osvix = svix;
2cf2cfc6 9368 bool is_utf8 = FALSE; /* is this item utf8? */
20f6aaab
AS
9369#ifdef HAS_LDBL_SPRINTF_BUG
9370 /* This is to try to fix a bug with irix/nonstop-ux/powerux and
205f51d8 9371 with sfio - Allen <allens@cpan.org> */
20f6aaab
AS
9372 bool fix_ldbl_sprintf_bug = FALSE;
9373#endif
205f51d8 9374
46fc3d4c 9375 char esignbuf[4];
89ebb4a3 9376 U8 utf8buf[UTF8_MAXBYTES+1];
46fc3d4c 9377 STRLEN esignlen = 0;
9378
bd61b366 9379 const char *eptr = NULL;
1d1ac7bc 9380 const char *fmtstart;
fc36a67e 9381 STRLEN elen = 0;
a0714e2c 9382 SV *vecsv = NULL;
4608196e 9383 const U8 *vecstr = NULL;
b22c7a20 9384 STRLEN veclen = 0;
934abaf1 9385 char c = 0;
46fc3d4c 9386 int i;
9c5ffd7c 9387 unsigned base = 0;
8c8eb53c
RB
9388 IV iv = 0;
9389 UV uv = 0;
9e5b023a
JH
9390 /* we need a long double target in case HAS_LONG_DOUBLE but
9391 not USE_LONG_DOUBLE
9392 */
35fff930 9393#if defined(HAS_LONG_DOUBLE) && LONG_DOUBLESIZE > DOUBLESIZE
9e5b023a
JH
9394 long double nv;
9395#else
65202027 9396 NV nv;
9e5b023a 9397#endif
46fc3d4c 9398 STRLEN have;
9399 STRLEN need;
9400 STRLEN gap;
7af36d83 9401 const char *dotstr = ".";
b22c7a20 9402 STRLEN dotstrlen = 1;
211dfcf1 9403 I32 efix = 0; /* explicit format parameter index */
eb3fce90 9404 I32 ewix = 0; /* explicit width index */
211dfcf1
HS
9405 I32 epix = 0; /* explicit precision index */
9406 I32 evix = 0; /* explicit vector index */
eb3fce90 9407 bool asterisk = FALSE;
46fc3d4c 9408
211dfcf1 9409 /* echo everything up to the next format specification */
46fc3d4c 9410 for (q = p; q < patend && *q != '%'; ++q) ;
9411 if (q > p) {
db79b45b
JH
9412 if (has_utf8 && !pat_utf8)
9413 sv_catpvn_utf8_upgrade(sv, p, q - p, nsv);
9414 else
9415 sv_catpvn(sv, p, q - p);
46fc3d4c 9416 p = q;
9417 }
9418 if (q++ >= patend)
9419 break;
9420
1d1ac7bc
MHM
9421 fmtstart = q;
9422
211dfcf1
HS
9423/*
9424 We allow format specification elements in this order:
9425 \d+\$ explicit format parameter index
9426 [-+ 0#]+ flags
a472f209 9427 v|\*(\d+\$)?v vector with optional (optionally specified) arg
f3583277 9428 0 flag (as above): repeated to allow "v02"
211dfcf1
HS
9429 \d+|\*(\d+\$)? width using optional (optionally specified) arg
9430 \.(\d*|\*(\d+\$)?) precision using optional (optionally specified) arg
9431 [hlqLV] size
8896765a
RB
9432 [%bcdefginopsuxDFOUX] format (mandatory)
9433*/
9434
9435 if (args) {
9436/*
9437 As of perl5.9.3, printf format checking is on by default.
9438 Internally, perl uses %p formats to provide an escape to
9439 some extended formatting. This block deals with those
9440 extensions: if it does not match, (char*)q is reset and
9441 the normal format processing code is used.
9442
9443 Currently defined extensions are:
9444 %p include pointer address (standard)
9445 %-p (SVf) include an SV (previously %_)
9446 %-<num>p include an SV with precision <num>
8896765a
RB
9447 %<num>p reserved for future extensions
9448
9449 Robin Barker 2005-07-14
f46d31f2
RB
9450
9451 %1p (VDf) removed. RMB 2007-10-19
211dfcf1 9452*/
8896765a
RB
9453 char* r = q;
9454 bool sv = FALSE;
9455 STRLEN n = 0;
9456 if (*q == '-')
9457 sv = *q++;
c445ea15 9458 n = expect_number(&q);
8896765a
RB
9459 if (*q++ == 'p') {
9460 if (sv) { /* SVf */
9461 if (n) {
9462 precis = n;
9463 has_precis = TRUE;
9464 }
daba3364 9465 argsv = MUTABLE_SV(va_arg(*args, void*));
4ea561bc 9466 eptr = SvPV_const(argsv, elen);
8896765a
RB
9467 if (DO_UTF8(argsv))
9468 is_utf8 = TRUE;
9469 goto string;
9470 }
8896765a 9471 else if (n) {
9b387841
NC
9472 Perl_ck_warner_d(aTHX_ packWARN(WARN_INTERNAL),
9473 "internal %%<num>p might conflict with future printf extensions");
8896765a
RB
9474 }
9475 }
9476 q = r;
9477 }
9478
c445ea15 9479 if ( (width = expect_number(&q)) ) {
211dfcf1
HS
9480 if (*q == '$') {
9481 ++q;
9482 efix = width;
9483 } else {
9484 goto gotwidth;
9485 }
9486 }
9487
fc36a67e 9488 /* FLAGS */
9489
46fc3d4c 9490 while (*q) {
9491 switch (*q) {
9492 case ' ':
9493 case '+':
9911cee9
TS
9494 if (plus == '+' && *q == ' ') /* '+' over ' ' */
9495 q++;
9496 else
9497 plus = *q++;
46fc3d4c 9498 continue;
9499
9500 case '-':
9501 left = TRUE;
9502 q++;
9503 continue;
9504
9505 case '0':
9506 fill = *q++;
9507 continue;
9508
9509 case '#':
9510 alt = TRUE;
9511 q++;
9512 continue;
9513
fc36a67e 9514 default:
9515 break;
9516 }
9517 break;
9518 }
46fc3d4c 9519
211dfcf1 9520 tryasterisk:
eb3fce90 9521 if (*q == '*') {
211dfcf1 9522 q++;
c445ea15 9523 if ( (ewix = expect_number(&q)) )
211dfcf1
HS
9524 if (*q++ != '$')
9525 goto unknown;
eb3fce90 9526 asterisk = TRUE;
211dfcf1
HS
9527 }
9528 if (*q == 'v') {
eb3fce90 9529 q++;
211dfcf1
HS
9530 if (vectorize)
9531 goto unknown;
9cbac4c7 9532 if ((vectorarg = asterisk)) {
211dfcf1
HS
9533 evix = ewix;
9534 ewix = 0;
9535 asterisk = FALSE;
9536 }
9537 vectorize = TRUE;
9538 goto tryasterisk;
eb3fce90
JH
9539 }
9540
211dfcf1 9541 if (!asterisk)
858a90f9 9542 {
7a5fa8a2 9543 if( *q == '0' )
f3583277 9544 fill = *q++;
c445ea15 9545 width = expect_number(&q);
858a90f9 9546 }
211dfcf1
HS
9547
9548 if (vectorize) {
9549 if (vectorarg) {
9550 if (args)
9551 vecsv = va_arg(*args, SV*);
7ad96abb
NC
9552 else if (evix) {
9553 vecsv = (evix > 0 && evix <= svmax)
81ae3cde 9554 ? svargs[evix-1] : S_vcatpvfn_missing_argument(aTHX);
7ad96abb 9555 } else {
7baa4690 9556 vecsv = svix < svmax
81ae3cde 9557 ? svargs[svix++] : S_vcatpvfn_missing_argument(aTHX);
7ad96abb 9558 }
245d4a47 9559 dotstr = SvPV_const(vecsv, dotstrlen);
640283f5
NC
9560 /* Keep the DO_UTF8 test *after* the SvPV call, else things go
9561 bad with tied or overloaded values that return UTF8. */
211dfcf1 9562 if (DO_UTF8(vecsv))
2cf2cfc6 9563 is_utf8 = TRUE;
640283f5
NC
9564 else if (has_utf8) {
9565 vecsv = sv_mortalcopy(vecsv);
9566 sv_utf8_upgrade(vecsv);
9567 dotstr = SvPV_const(vecsv, dotstrlen);
9568 is_utf8 = TRUE;
9569 }
211dfcf1
HS
9570 }
9571 if (args) {
8896765a 9572 VECTORIZE_ARGS
eb3fce90 9573 }
7ad96abb 9574 else if (efix ? (efix > 0 && efix <= svmax) : svix < svmax) {
211dfcf1 9575 vecsv = svargs[efix ? efix-1 : svix++];
245d4a47 9576 vecstr = (U8*)SvPV_const(vecsv,veclen);
2cf2cfc6 9577 vec_utf8 = DO_UTF8(vecsv);
96b8f7ce
JP
9578
9579 /* if this is a version object, we need to convert
9580 * back into v-string notation and then let the
9581 * vectorize happen normally
d7aa5382 9582 */
96b8f7ce
JP
9583 if (sv_derived_from(vecsv, "version")) {
9584 char *version = savesvpv(vecsv);
85fbaab2 9585 if ( hv_exists(MUTABLE_HV(SvRV(vecsv)), "alpha", 5 ) ) {
34ba6322
SP
9586 Perl_warner(aTHX_ packWARN(WARN_INTERNAL),
9587 "vector argument not supported with alpha versions");
9588 goto unknown;
9589 }
96b8f7ce 9590 vecsv = sv_newmortal();
65b06e02 9591 scan_vstring(version, version + veclen, vecsv);
96b8f7ce
JP
9592 vecstr = (U8*)SvPV_const(vecsv, veclen);
9593 vec_utf8 = DO_UTF8(vecsv);
9594 Safefree(version);
d7aa5382 9595 }
211dfcf1
HS
9596 }
9597 else {
9598 vecstr = (U8*)"";
9599 veclen = 0;
9600 }
eb3fce90 9601 }
fc36a67e 9602
eb3fce90 9603 if (asterisk) {
fc36a67e 9604 if (args)
9605 i = va_arg(*args, int);
9606 else
eb3fce90
JH
9607 i = (ewix ? ewix <= svmax : svix < svmax) ?
9608 SvIVx(svargs[ewix ? ewix-1 : svix++]) : 0;
fc36a67e 9609 left |= (i < 0);
9610 width = (i < 0) ? -i : i;
fc36a67e 9611 }
211dfcf1 9612 gotwidth:
fc36a67e 9613
9614 /* PRECISION */
46fc3d4c 9615
fc36a67e 9616 if (*q == '.') {
9617 q++;
9618 if (*q == '*') {
211dfcf1 9619 q++;
c445ea15 9620 if ( ((epix = expect_number(&q))) && (*q++ != '$') )
7b8dd722
HS
9621 goto unknown;
9622 /* XXX: todo, support specified precision parameter */
9623 if (epix)
211dfcf1 9624 goto unknown;
46fc3d4c 9625 if (args)
9626 i = va_arg(*args, int);
9627 else
eb3fce90
JH
9628 i = (ewix ? ewix <= svmax : svix < svmax)
9629 ? SvIVx(svargs[ewix ? ewix-1 : svix++]) : 0;
9911cee9
TS
9630 precis = i;
9631 has_precis = !(i < 0);
fc36a67e 9632 }
9633 else {
9634 precis = 0;
9635 while (isDIGIT(*q))
9636 precis = precis * 10 + (*q++ - '0');
9911cee9 9637 has_precis = TRUE;
fc36a67e 9638 }
fc36a67e 9639 }
46fc3d4c 9640
fc36a67e 9641 /* SIZE */
46fc3d4c 9642
fc36a67e 9643 switch (*q) {
c623ac67
GS
9644#ifdef WIN32
9645 case 'I': /* Ix, I32x, and I64x */
9646# ifdef WIN64
9647 if (q[1] == '6' && q[2] == '4') {
9648 q += 3;
9649 intsize = 'q';
9650 break;
9651 }
9652# endif
9653 if (q[1] == '3' && q[2] == '2') {
9654 q += 3;
9655 break;
9656 }
9657# ifdef WIN64
9658 intsize = 'q';
9659# endif
9660 q++;
9661 break;
9662#endif
9e5b023a 9663#if defined(HAS_QUAD) || defined(HAS_LONG_DOUBLE)
6f9bb7fd 9664 case 'L': /* Ld */
5f66b61c 9665 /*FALLTHROUGH*/
e5c81feb 9666#ifdef HAS_QUAD
6f9bb7fd 9667 case 'q': /* qd */
9e5b023a 9668#endif
6f9bb7fd
GS
9669 intsize = 'q';
9670 q++;
9671 break;
9672#endif
fc36a67e 9673 case 'l':
9e5b023a 9674#if defined(HAS_QUAD) || defined(HAS_LONG_DOUBLE)
205f51d8 9675 if (*(q + 1) == 'l') { /* lld, llf */
fc36a67e 9676 intsize = 'q';
9677 q += 2;
46fc3d4c 9678 break;
cf2093f6 9679 }
fc36a67e 9680#endif
5f66b61c 9681 /*FALLTHROUGH*/
fc36a67e 9682 case 'h':
5f66b61c 9683 /*FALLTHROUGH*/
fc36a67e 9684 case 'V':
9685 intsize = *q++;
46fc3d4c 9686 break;
9687 }
9688
fc36a67e 9689 /* CONVERSION */
9690
211dfcf1
HS
9691 if (*q == '%') {
9692 eptr = q++;
9693 elen = 1;
26372e71
GA
9694 if (vectorize) {
9695 c = '%';
9696 goto unknown;
9697 }
211dfcf1
HS
9698 goto string;
9699 }
9700
26372e71 9701 if (!vectorize && !args) {
86c51f8b
NC
9702 if (efix) {
9703 const I32 i = efix-1;
7baa4690 9704 argsv = (i >= 0 && i < svmax)
81ae3cde 9705 ? svargs[i] : S_vcatpvfn_missing_argument(aTHX);
86c51f8b
NC
9706 } else {
9707 argsv = (svix >= 0 && svix < svmax)
81ae3cde 9708 ? svargs[svix++] : S_vcatpvfn_missing_argument(aTHX);
86c51f8b 9709 }
863811b2 9710 }
211dfcf1 9711
46fc3d4c 9712 switch (c = *q++) {
9713
9714 /* STRINGS */
9715
46fc3d4c 9716 case 'c':
26372e71
GA
9717 if (vectorize)
9718 goto unknown;
4ea561bc 9719 uv = (args) ? va_arg(*args, int) : SvIV(argsv);
1bd104fb
JH
9720 if ((uv > 255 ||
9721 (!UNI_IS_INVARIANT(uv) && SvUTF8(sv)))
0064a8a9 9722 && !IN_BYTES) {
dfe13c55 9723 eptr = (char*)utf8buf;
9041c2e3 9724 elen = uvchr_to_utf8((U8*)eptr, uv) - utf8buf;
2cf2cfc6 9725 is_utf8 = TRUE;
7e2040f0
GS
9726 }
9727 else {
9728 c = (char)uv;
9729 eptr = &c;
9730 elen = 1;
a0ed51b3 9731 }
46fc3d4c 9732 goto string;
9733
46fc3d4c 9734 case 's':
26372e71
GA
9735 if (vectorize)
9736 goto unknown;
9737 if (args) {
fc36a67e 9738 eptr = va_arg(*args, char*);
c635e13b 9739 if (eptr)
9740 elen = strlen(eptr);
9741 else {
27da23d5 9742 eptr = (char *)nullstr;
c635e13b 9743 elen = sizeof nullstr - 1;
9744 }
46fc3d4c 9745 }
211dfcf1 9746 else {
4ea561bc 9747 eptr = SvPV_const(argsv, elen);
7e2040f0 9748 if (DO_UTF8(argsv)) {
c494f1f4 9749 STRLEN old_precis = precis;
a0ed51b3 9750 if (has_precis && precis < elen) {
c494f1f4 9751 STRLEN ulen = sv_len_utf8(argsv);
9ef5ed94 9752 I32 p = precis > ulen ? ulen : precis;
7e2040f0 9753 sv_pos_u2b(argsv, &p, 0); /* sticks at end */
a0ed51b3
LW
9754 precis = p;
9755 }
9756 if (width) { /* fudge width (can't fudge elen) */
59b61096
AV
9757 if (has_precis && precis < elen)
9758 width += precis - old_precis;
9759 else
9760 width += elen - sv_len_utf8(argsv);
a0ed51b3 9761 }
2cf2cfc6 9762 is_utf8 = TRUE;
a0ed51b3
LW
9763 }
9764 }
fc36a67e 9765
46fc3d4c 9766 string:
9ef5ed94 9767 if (has_precis && precis < elen)
46fc3d4c 9768 elen = precis;
9769 break;
9770
9771 /* INTEGERS */
9772
fc36a67e 9773 case 'p':
be75b157 9774 if (alt || vectorize)
c2e66d9e 9775 goto unknown;
211dfcf1 9776 uv = PTR2UV(args ? va_arg(*args, void*) : argsv);
fc36a67e 9777 base = 16;
9778 goto integer;
9779
46fc3d4c 9780 case 'D':
29fe7a80 9781#ifdef IV_IS_QUAD
22f3ae8c 9782 intsize = 'q';
29fe7a80 9783#else
46fc3d4c 9784 intsize = 'l';
29fe7a80 9785#endif
5f66b61c 9786 /*FALLTHROUGH*/
46fc3d4c 9787 case 'd':
9788 case 'i':
8896765a
RB
9789#if vdNUMBER
9790 format_vd:
9791#endif
b22c7a20 9792 if (vectorize) {
ba210ebe 9793 STRLEN ulen;
211dfcf1
HS
9794 if (!veclen)
9795 continue;
2cf2cfc6
A
9796 if (vec_utf8)
9797 uv = utf8n_to_uvchr(vecstr, veclen, &ulen,
9798 UTF8_ALLOW_ANYUV);
b22c7a20 9799 else {
e83d50c9 9800 uv = *vecstr;
b22c7a20
GS
9801 ulen = 1;
9802 }
9803 vecstr += ulen;
9804 veclen -= ulen;
e83d50c9
JP
9805 if (plus)
9806 esignbuf[esignlen++] = plus;
b22c7a20
GS
9807 }
9808 else if (args) {
46fc3d4c 9809 switch (intsize) {
9810 case 'h': iv = (short)va_arg(*args, int); break;
46fc3d4c 9811 case 'l': iv = va_arg(*args, long); break;
fc36a67e 9812 case 'V': iv = va_arg(*args, IV); break;
b10c0dba 9813 default: iv = va_arg(*args, int); break;
53f65a9e 9814 case 'q':
cf2093f6 9815#ifdef HAS_QUAD
53f65a9e
HS
9816 iv = va_arg(*args, Quad_t); break;
9817#else
9818 goto unknown;
cf2093f6 9819#endif
46fc3d4c 9820 }
9821 }
9822 else {
4ea561bc 9823 IV tiv = SvIV(argsv); /* work around GCC bug #13488 */
46fc3d4c 9824 switch (intsize) {
b10c0dba
MHM
9825 case 'h': iv = (short)tiv; break;
9826 case 'l': iv = (long)tiv; break;
9827 case 'V':
9828 default: iv = tiv; break;
53f65a9e 9829 case 'q':
cf2093f6 9830#ifdef HAS_QUAD
53f65a9e
HS
9831 iv = (Quad_t)tiv; break;
9832#else
9833 goto unknown;
cf2093f6 9834#endif
46fc3d4c 9835 }
9836 }
e83d50c9
JP
9837 if ( !vectorize ) /* we already set uv above */
9838 {
9839 if (iv >= 0) {
9840 uv = iv;
9841 if (plus)
9842 esignbuf[esignlen++] = plus;
9843 }
9844 else {
9845 uv = -iv;
9846 esignbuf[esignlen++] = '-';
9847 }
46fc3d4c 9848 }
9849 base = 10;
9850 goto integer;
9851
fc36a67e 9852 case 'U':
29fe7a80 9853#ifdef IV_IS_QUAD
22f3ae8c 9854 intsize = 'q';
29fe7a80 9855#else
fc36a67e 9856 intsize = 'l';
29fe7a80 9857#endif
5f66b61c 9858 /*FALLTHROUGH*/
fc36a67e 9859 case 'u':
9860 base = 10;
9861 goto uns_integer;
9862
7ff06cc7 9863 case 'B':
4f19785b
WSI
9864 case 'b':
9865 base = 2;
9866 goto uns_integer;
9867
46fc3d4c 9868 case 'O':
29fe7a80 9869#ifdef IV_IS_QUAD
22f3ae8c 9870 intsize = 'q';
29fe7a80 9871#else
46fc3d4c 9872 intsize = 'l';
29fe7a80 9873#endif
5f66b61c 9874 /*FALLTHROUGH*/
46fc3d4c 9875 case 'o':
9876 base = 8;
9877 goto uns_integer;
9878
9879 case 'X':
46fc3d4c 9880 case 'x':
9881 base = 16;
46fc3d4c 9882
9883 uns_integer:
b22c7a20 9884 if (vectorize) {
ba210ebe 9885 STRLEN ulen;
b22c7a20 9886 vector:
211dfcf1
HS
9887 if (!veclen)
9888 continue;
2cf2cfc6
A
9889 if (vec_utf8)
9890 uv = utf8n_to_uvchr(vecstr, veclen, &ulen,
9891 UTF8_ALLOW_ANYUV);
b22c7a20 9892 else {
a05b299f 9893 uv = *vecstr;
b22c7a20
GS
9894 ulen = 1;
9895 }
9896 vecstr += ulen;
9897 veclen -= ulen;
9898 }
9899 else if (args) {
46fc3d4c 9900 switch (intsize) {
9901 case 'h': uv = (unsigned short)va_arg(*args, unsigned); break;
46fc3d4c 9902 case 'l': uv = va_arg(*args, unsigned long); break;
fc36a67e 9903 case 'V': uv = va_arg(*args, UV); break;
b10c0dba 9904 default: uv = va_arg(*args, unsigned); break;
53f65a9e 9905 case 'q':
cf2093f6 9906#ifdef HAS_QUAD
53f65a9e
HS
9907 uv = va_arg(*args, Uquad_t); break;
9908#else
9909 goto unknown;
cf2093f6 9910#endif
46fc3d4c 9911 }
9912 }
9913 else {
4ea561bc 9914 UV tuv = SvUV(argsv); /* work around GCC bug #13488 */
46fc3d4c 9915 switch (intsize) {
b10c0dba
MHM
9916 case 'h': uv = (unsigned short)tuv; break;
9917 case 'l': uv = (unsigned long)tuv; break;
9918 case 'V':
9919 default: uv = tuv; break;
53f65a9e 9920 case 'q':
cf2093f6 9921#ifdef HAS_QUAD
53f65a9e
HS
9922 uv = (Uquad_t)tuv; break;
9923#else
9924 goto unknown;
cf2093f6 9925#endif
46fc3d4c 9926 }
9927 }
9928
9929 integer:
4d84ee25
NC
9930 {
9931 char *ptr = ebuf + sizeof ebuf;
1387f30c
DD
9932 bool tempalt = uv ? alt : FALSE; /* Vectors can't change alt */
9933 zeros = 0;
9934
4d84ee25
NC
9935 switch (base) {
9936 unsigned dig;
9937 case 16:
14eb61ab 9938 p = (char *)((c == 'X') ? PL_hexdigit + 16 : PL_hexdigit);
4d84ee25
NC
9939 do {
9940 dig = uv & 15;
9941 *--ptr = p[dig];
9942 } while (uv >>= 4);
1387f30c 9943 if (tempalt) {
4d84ee25
NC
9944 esignbuf[esignlen++] = '0';
9945 esignbuf[esignlen++] = c; /* 'x' or 'X' */
9946 }
9947 break;
9948 case 8:
9949 do {
9950 dig = uv & 7;
9951 *--ptr = '0' + dig;
9952 } while (uv >>= 3);
9953 if (alt && *ptr != '0')
9954 *--ptr = '0';
9955 break;
9956 case 2:
9957 do {
9958 dig = uv & 1;
9959 *--ptr = '0' + dig;
9960 } while (uv >>= 1);
1387f30c 9961 if (tempalt) {
4d84ee25 9962 esignbuf[esignlen++] = '0';
7ff06cc7 9963 esignbuf[esignlen++] = c;
4d84ee25
NC
9964 }
9965 break;
9966 default: /* it had better be ten or less */
9967 do {
9968 dig = uv % base;
9969 *--ptr = '0' + dig;
9970 } while (uv /= base);
9971 break;
46fc3d4c 9972 }
4d84ee25
NC
9973 elen = (ebuf + sizeof ebuf) - ptr;
9974 eptr = ptr;
9975 if (has_precis) {
9976 if (precis > elen)
9977 zeros = precis - elen;
e6bb52fd
TS
9978 else if (precis == 0 && elen == 1 && *eptr == '0'
9979 && !(base == 8 && alt)) /* "%#.0o" prints "0" */
4d84ee25 9980 elen = 0;
9911cee9
TS
9981
9982 /* a precision nullifies the 0 flag. */
9983 if (fill == '0')
9984 fill = ' ';
eda88b6d 9985 }
c10ed8b9 9986 }
46fc3d4c 9987 break;
9988
9989 /* FLOATING POINT */
9990
fc36a67e 9991 case 'F':
9992 c = 'f'; /* maybe %F isn't supported here */
5f66b61c 9993 /*FALLTHROUGH*/
46fc3d4c 9994 case 'e': case 'E':
fc36a67e 9995 case 'f':
46fc3d4c 9996 case 'g': case 'G':
26372e71
GA
9997 if (vectorize)
9998 goto unknown;
46fc3d4c 9999
10000 /* This is evil, but floating point is even more evil */
10001
9e5b023a
JH
10002 /* for SV-style calling, we can only get NV
10003 for C-style calling, we assume %f is double;
10004 for simplicity we allow any of %Lf, %llf, %qf for long double
10005 */
10006 switch (intsize) {
10007 case 'V':
10008#if defined(USE_LONG_DOUBLE)
10009 intsize = 'q';
10010#endif
10011 break;
8a2e3f14 10012/* [perl #20339] - we should accept and ignore %lf rather than die */
00e17364 10013 case 'l':
5f66b61c 10014 /*FALLTHROUGH*/
9e5b023a
JH
10015 default:
10016#if defined(USE_LONG_DOUBLE)
10017 intsize = args ? 0 : 'q';
10018#endif
10019 break;
10020 case 'q':
10021#if defined(HAS_LONG_DOUBLE)
10022 break;
10023#else
5f66b61c 10024 /*FALLTHROUGH*/
9e5b023a
JH
10025#endif
10026 case 'h':
9e5b023a
JH
10027 goto unknown;
10028 }
10029
10030 /* now we need (long double) if intsize == 'q', else (double) */
26372e71 10031 nv = (args) ?
35fff930
JH
10032#if LONG_DOUBLESIZE > DOUBLESIZE
10033 intsize == 'q' ?
205f51d8
AS
10034 va_arg(*args, long double) :
10035 va_arg(*args, double)
35fff930 10036#else
205f51d8 10037 va_arg(*args, double)
35fff930 10038#endif
4ea561bc 10039 : SvNV(argsv);
fc36a67e 10040
10041 need = 0;
3952c29a
NC
10042 /* nv * 0 will be NaN for NaN, +Inf and -Inf, and 0 for anything
10043 else. frexp() has some unspecified behaviour for those three */
10044 if (c != 'e' && c != 'E' && (nv * 0) == 0) {
fc36a67e 10045 i = PERL_INT_MIN;
9e5b023a
JH
10046 /* FIXME: if HAS_LONG_DOUBLE but not USE_LONG_DOUBLE this
10047 will cast our (long double) to (double) */
73b309ea 10048 (void)Perl_frexp(nv, &i);
fc36a67e 10049 if (i == PERL_INT_MIN)
cea2e8a9 10050 Perl_die(aTHX_ "panic: frexp");
c635e13b 10051 if (i > 0)
fc36a67e 10052 need = BIT_DIGITS(i);
10053 }
10054 need += has_precis ? precis : 6; /* known default */
20f6aaab 10055
fc36a67e 10056 if (need < width)
10057 need = width;
10058
20f6aaab
AS
10059#ifdef HAS_LDBL_SPRINTF_BUG
10060 /* This is to try to fix a bug with irix/nonstop-ux/powerux and
205f51d8
AS
10061 with sfio - Allen <allens@cpan.org> */
10062
10063# ifdef DBL_MAX
10064# define MY_DBL_MAX DBL_MAX
10065# else /* XXX guessing! HUGE_VAL may be defined as infinity, so not using */
10066# if DOUBLESIZE >= 8
10067# define MY_DBL_MAX 1.7976931348623157E+308L
10068# else
10069# define MY_DBL_MAX 3.40282347E+38L
10070# endif
10071# endif
10072
10073# ifdef HAS_LDBL_SPRINTF_BUG_LESS1 /* only between -1L & 1L - Allen */
10074# define MY_DBL_MAX_BUG 1L
20f6aaab 10075# else
205f51d8 10076# define MY_DBL_MAX_BUG MY_DBL_MAX
20f6aaab 10077# endif
20f6aaab 10078
205f51d8
AS
10079# ifdef DBL_MIN
10080# define MY_DBL_MIN DBL_MIN
10081# else /* XXX guessing! -Allen */
10082# if DOUBLESIZE >= 8
10083# define MY_DBL_MIN 2.2250738585072014E-308L
10084# else
10085# define MY_DBL_MIN 1.17549435E-38L
10086# endif
10087# endif
20f6aaab 10088
205f51d8
AS
10089 if ((intsize == 'q') && (c == 'f') &&
10090 ((nv < MY_DBL_MAX_BUG) && (nv > -MY_DBL_MAX_BUG)) &&
10091 (need < DBL_DIG)) {
10092 /* it's going to be short enough that
10093 * long double precision is not needed */
10094
10095 if ((nv <= 0L) && (nv >= -0L))
10096 fix_ldbl_sprintf_bug = TRUE; /* 0 is 0 - easiest */
10097 else {
10098 /* would use Perl_fp_class as a double-check but not
10099 * functional on IRIX - see perl.h comments */
10100
10101 if ((nv >= MY_DBL_MIN) || (nv <= -MY_DBL_MIN)) {
10102 /* It's within the range that a double can represent */
10103#if defined(DBL_MAX) && !defined(DBL_MIN)
10104 if ((nv >= ((long double)1/DBL_MAX)) ||
10105 (nv <= (-(long double)1/DBL_MAX)))
20f6aaab 10106#endif
205f51d8 10107 fix_ldbl_sprintf_bug = TRUE;
20f6aaab 10108 }
205f51d8
AS
10109 }
10110 if (fix_ldbl_sprintf_bug == TRUE) {
10111 double temp;
10112
10113 intsize = 0;
10114 temp = (double)nv;
10115 nv = (NV)temp;
10116 }
20f6aaab 10117 }
205f51d8
AS
10118
10119# undef MY_DBL_MAX
10120# undef MY_DBL_MAX_BUG
10121# undef MY_DBL_MIN
10122
20f6aaab
AS
10123#endif /* HAS_LDBL_SPRINTF_BUG */
10124
46fc3d4c 10125 need += 20; /* fudge factor */
80252599
GS
10126 if (PL_efloatsize < need) {
10127 Safefree(PL_efloatbuf);
10128 PL_efloatsize = need + 20; /* more fudge */
a02a5408 10129 Newx(PL_efloatbuf, PL_efloatsize, char);
7d5ea4e7 10130 PL_efloatbuf[0] = '\0';
46fc3d4c 10131 }
10132
4151a5fe
IZ
10133 if ( !(width || left || plus || alt) && fill != '0'
10134 && has_precis && intsize != 'q' ) { /* Shortcuts */
2873255c
NC
10135 /* See earlier comment about buggy Gconvert when digits,
10136 aka precis is 0 */
10137 if ( c == 'g' && precis) {
2e59c212 10138 Gconvert((NV)nv, (int)precis, 0, PL_efloatbuf);
4150c189
NC
10139 /* May return an empty string for digits==0 */
10140 if (*PL_efloatbuf) {
10141 elen = strlen(PL_efloatbuf);
4151a5fe 10142 goto float_converted;
4150c189 10143 }
4151a5fe
IZ
10144 } else if ( c == 'f' && !precis) {
10145 if ((eptr = F0convert(nv, ebuf + sizeof ebuf, &elen)))
10146 break;
10147 }
10148 }
4d84ee25
NC
10149 {
10150 char *ptr = ebuf + sizeof ebuf;
10151 *--ptr = '\0';
10152 *--ptr = c;
10153 /* FIXME: what to do if HAS_LONG_DOUBLE but not PERL_PRIfldbl? */
9e5b023a 10154#if defined(HAS_LONG_DOUBLE) && defined(PERL_PRIfldbl)
4d84ee25
NC
10155 if (intsize == 'q') {
10156 /* Copy the one or more characters in a long double
10157 * format before the 'base' ([efgEFG]) character to
10158 * the format string. */
10159 static char const prifldbl[] = PERL_PRIfldbl;
10160 char const *p = prifldbl + sizeof(prifldbl) - 3;
10161 while (p >= prifldbl) { *--ptr = *p--; }
10162 }
65202027 10163#endif
4d84ee25
NC
10164 if (has_precis) {
10165 base = precis;
10166 do { *--ptr = '0' + (base % 10); } while (base /= 10);
10167 *--ptr = '.';
10168 }
10169 if (width) {
10170 base = width;
10171 do { *--ptr = '0' + (base % 10); } while (base /= 10);
10172 }
10173 if (fill == '0')
10174 *--ptr = fill;
10175 if (left)
10176 *--ptr = '-';
10177 if (plus)
10178 *--ptr = plus;
10179 if (alt)
10180 *--ptr = '#';
10181 *--ptr = '%';
10182
10183 /* No taint. Otherwise we are in the strange situation
10184 * where printf() taints but print($float) doesn't.
10185 * --jhi */
9e5b023a 10186#if defined(HAS_LONG_DOUBLE)
4150c189 10187 elen = ((intsize == 'q')
d9fad198
JH
10188 ? my_snprintf(PL_efloatbuf, PL_efloatsize, ptr, nv)
10189 : my_snprintf(PL_efloatbuf, PL_efloatsize, ptr, (double)nv));
9e5b023a 10190#else
4150c189 10191 elen = my_sprintf(PL_efloatbuf, ptr, nv);
9e5b023a 10192#endif
4d84ee25 10193 }
4151a5fe 10194 float_converted:
80252599 10195 eptr = PL_efloatbuf;
46fc3d4c 10196 break;
10197
fc36a67e 10198 /* SPECIAL */
10199
10200 case 'n':
26372e71
GA
10201 if (vectorize)
10202 goto unknown;
fc36a67e 10203 i = SvCUR(sv) - origlen;
26372e71 10204 if (args) {
c635e13b 10205 switch (intsize) {
10206 case 'h': *(va_arg(*args, short*)) = i; break;
10207 default: *(va_arg(*args, int*)) = i; break;
10208 case 'l': *(va_arg(*args, long*)) = i; break;
10209 case 'V': *(va_arg(*args, IV*)) = i; break;
53f65a9e 10210 case 'q':
cf2093f6 10211#ifdef HAS_QUAD
53f65a9e
HS
10212 *(va_arg(*args, Quad_t*)) = i; break;
10213#else
10214 goto unknown;
cf2093f6 10215#endif
c635e13b 10216 }
fc36a67e 10217 }
9dd79c3f 10218 else
211dfcf1 10219 sv_setuv_mg(argsv, (UV)i);
fc36a67e 10220 continue; /* not "break" */
10221
10222 /* UNKNOWN */
10223
46fc3d4c 10224 default:
fc36a67e 10225 unknown:
041457d9
DM
10226 if (!args
10227 && (PL_op->op_type == OP_PRTF || PL_op->op_type == OP_SPRINTF)
10228 && ckWARN(WARN_PRINTF))
10229 {
c4420975 10230 SV * const msg = sv_newmortal();
35c1215d
NC
10231 Perl_sv_setpvf(aTHX_ msg, "Invalid conversion in %sprintf: ",
10232 (PL_op->op_type == OP_PRTF) ? "" : "s");
1d1ac7bc
MHM
10233 if (fmtstart < patend) {
10234 const char * const fmtend = q < patend ? q : patend;
10235 const char * f;
10236 sv_catpvs(msg, "\"%");
10237 for (f = fmtstart; f < fmtend; f++) {
10238 if (isPRINT(*f)) {
10239 sv_catpvn(msg, f, 1);
10240 } else {
10241 Perl_sv_catpvf(aTHX_ msg,
10242 "\\%03"UVof, (UV)*f & 0xFF);
10243 }
10244 }
10245 sv_catpvs(msg, "\"");
10246 } else {
396482e1 10247 sv_catpvs(msg, "end of string");
1d1ac7bc 10248 }
be2597df 10249 Perl_warner(aTHX_ packWARN(WARN_PRINTF), "%"SVf, SVfARG(msg)); /* yes, this is reentrant */
c635e13b 10250 }
fb73857a 10251
10252 /* output mangled stuff ... */
10253 if (c == '\0')
10254 --q;
46fc3d4c 10255 eptr = p;
10256 elen = q - p;
fb73857a 10257
10258 /* ... right here, because formatting flags should not apply */
10259 SvGROW(sv, SvCUR(sv) + elen + 1);
10260 p = SvEND(sv);
4459522c 10261 Copy(eptr, p, elen, char);
fb73857a 10262 p += elen;
10263 *p = '\0';
3f7c398e 10264 SvCUR_set(sv, p - SvPVX_const(sv));
58e33a90 10265 svix = osvix;
fb73857a 10266 continue; /* not "break" */
46fc3d4c 10267 }
10268
cc61b222
TS
10269 if (is_utf8 != has_utf8) {
10270 if (is_utf8) {
10271 if (SvCUR(sv))
10272 sv_utf8_upgrade(sv);
10273 }
10274 else {
10275 const STRLEN old_elen = elen;
59cd0e26 10276 SV * const nsv = newSVpvn_flags(eptr, elen, SVs_TEMP);
cc61b222
TS
10277 sv_utf8_upgrade(nsv);
10278 eptr = SvPVX_const(nsv);
10279 elen = SvCUR(nsv);
10280
10281 if (width) { /* fudge width (can't fudge elen) */
10282 width += elen - old_elen;
10283 }
10284 is_utf8 = TRUE;
10285 }
10286 }
10287
6c94ec8b 10288 have = esignlen + zeros + elen;
ed2b91d2 10289 if (have < zeros)
f1f66076 10290 Perl_croak_nocontext("%s", PL_memory_wrap);
6c94ec8b 10291
46fc3d4c 10292 need = (have > width ? have : width);
10293 gap = need - have;
10294
d2641cbd 10295 if (need >= (((STRLEN)~0) - SvCUR(sv) - dotstrlen - 1))
f1f66076 10296 Perl_croak_nocontext("%s", PL_memory_wrap);
b22c7a20 10297 SvGROW(sv, SvCUR(sv) + need + dotstrlen + 1);
46fc3d4c 10298 p = SvEND(sv);
10299 if (esignlen && fill == '0') {
53c1dcc0 10300 int i;
eb160463 10301 for (i = 0; i < (int)esignlen; i++)
46fc3d4c 10302 *p++ = esignbuf[i];
10303 }
10304 if (gap && !left) {
10305 memset(p, fill, gap);
10306 p += gap;
10307 }
10308 if (esignlen && fill != '0') {
53c1dcc0 10309 int i;
eb160463 10310 for (i = 0; i < (int)esignlen; i++)
46fc3d4c 10311 *p++ = esignbuf[i];
10312 }
fc36a67e 10313 if (zeros) {
53c1dcc0 10314 int i;
fc36a67e 10315 for (i = zeros; i; i--)
10316 *p++ = '0';
10317 }
46fc3d4c 10318 if (elen) {
4459522c 10319 Copy(eptr, p, elen, char);
46fc3d4c 10320 p += elen;
10321 }
10322 if (gap && left) {
10323 memset(p, ' ', gap);
10324 p += gap;
10325 }
b22c7a20
GS
10326 if (vectorize) {
10327 if (veclen) {
4459522c 10328 Copy(dotstr, p, dotstrlen, char);
b22c7a20
GS
10329 p += dotstrlen;
10330 }
10331 else
10332 vectorize = FALSE; /* done iterating over vecstr */
10333 }
2cf2cfc6
A
10334 if (is_utf8)
10335 has_utf8 = TRUE;
10336 if (has_utf8)
7e2040f0 10337 SvUTF8_on(sv);
46fc3d4c 10338 *p = '\0';
3f7c398e 10339 SvCUR_set(sv, p - SvPVX_const(sv));
b22c7a20
GS
10340 if (vectorize) {
10341 esignlen = 0;
10342 goto vector;
10343 }
46fc3d4c 10344 }
10345}
51371543 10346
645c22ef
DM
10347/* =========================================================================
10348
10349=head1 Cloning an interpreter
10350
10351All the macros and functions in this section are for the private use of
10352the main function, perl_clone().
10353
f2fc5c80 10354The foo_dup() functions make an exact copy of an existing foo thingy.
645c22ef
DM
10355During the course of a cloning, a hash table is used to map old addresses
10356to new addresses. The table is created and manipulated with the
10357ptr_table_* functions.
10358
10359=cut
10360
3e8320cc 10361 * =========================================================================*/
645c22ef
DM
10362
10363
1d7c1841
GS
10364#if defined(USE_ITHREADS)
10365
d4c19fe8 10366/* XXX Remove this so it doesn't have to go thru the macro and return for nothing */
1d7c1841
GS
10367#ifndef GpREFCNT_inc
10368# define GpREFCNT_inc(gp) ((gp) ? (++(gp)->gp_refcnt, (gp)) : (GP*)NULL)
10369#endif
10370
10371
a41cc44e 10372/* Certain cases in Perl_ss_dup have been merged, by relying on the fact
3e07292d 10373 that currently av_dup, gv_dup and hv_dup are the same as sv_dup.
538f2e76
NC
10374 If this changes, please unmerge ss_dup.
10375 Likewise, sv_dup_inc_multiple() relies on this fact. */
d2d73c3e 10376#define sv_dup_inc(s,t) SvREFCNT_inc(sv_dup(s,t))
7f466ec7 10377#define sv_dup_inc_NN(s,t) SvREFCNT_inc_NN(sv_dup(s,t))
502c6561
NC
10378#define av_dup(s,t) MUTABLE_AV(sv_dup((const SV *)s,t))
10379#define av_dup_inc(s,t) MUTABLE_AV(SvREFCNT_inc(sv_dup((const SV *)s,t)))
85fbaab2
NC
10380#define hv_dup(s,t) MUTABLE_HV(sv_dup((const SV *)s,t))
10381#define hv_dup_inc(s,t) MUTABLE_HV(SvREFCNT_inc(sv_dup((const SV *)s,t)))
daba3364 10382#define cv_dup(s,t) MUTABLE_CV(sv_dup((const SV *)s,t))
ea726b52 10383#define cv_dup_inc(s,t) MUTABLE_CV(SvREFCNT_inc(sv_dup((const SV *)s,t)))
daba3364 10384#define io_dup(s,t) MUTABLE_IO(sv_dup((const SV *)s,t))
a45c7426 10385#define io_dup_inc(s,t) MUTABLE_IO(SvREFCNT_inc(sv_dup((const SV *)s,t)))
159b6efe
NC
10386#define gv_dup(s,t) MUTABLE_GV(sv_dup((const SV *)s,t))
10387#define gv_dup_inc(s,t) MUTABLE_GV(SvREFCNT_inc(sv_dup((const SV *)s,t)))
6136c704
AL
10388#define SAVEPV(p) ((p) ? savepv(p) : NULL)
10389#define SAVEPVN(p,n) ((p) ? savepvn(p,n) : NULL)
8cf8f3d1 10390
199e78b7
DM
10391/* clone a parser */
10392
10393yy_parser *
66ceb532 10394Perl_parser_dup(pTHX_ const yy_parser *const proto, CLONE_PARAMS *const param)
199e78b7
DM
10395{
10396 yy_parser *parser;
10397
7918f24d
NC
10398 PERL_ARGS_ASSERT_PARSER_DUP;
10399
199e78b7
DM
10400 if (!proto)
10401 return NULL;
10402
7c197c94
DM
10403 /* look for it in the table first */
10404 parser = (yy_parser *)ptr_table_fetch(PL_ptr_table, proto);
10405 if (parser)
10406 return parser;
10407
10408 /* create anew and remember what it is */
199e78b7 10409 Newxz(parser, 1, yy_parser);
7c197c94 10410 ptr_table_store(PL_ptr_table, proto, parser);
199e78b7
DM
10411
10412 parser->yyerrstatus = 0;
10413 parser->yychar = YYEMPTY; /* Cause a token to be read. */
10414
10415 /* XXX these not yet duped */
10416 parser->old_parser = NULL;
10417 parser->stack = NULL;
10418 parser->ps = NULL;
10419 parser->stack_size = 0;
10420 /* XXX parser->stack->state = 0; */
10421
10422 /* XXX eventually, just Copy() most of the parser struct ? */
10423
10424 parser->lex_brackets = proto->lex_brackets;
10425 parser->lex_casemods = proto->lex_casemods;
10426 parser->lex_brackstack = savepvn(proto->lex_brackstack,
10427 (proto->lex_brackets < 120 ? 120 : proto->lex_brackets));
10428 parser->lex_casestack = savepvn(proto->lex_casestack,
10429 (proto->lex_casemods < 12 ? 12 : proto->lex_casemods));
10430 parser->lex_defer = proto->lex_defer;
10431 parser->lex_dojoin = proto->lex_dojoin;
10432 parser->lex_expect = proto->lex_expect;
10433 parser->lex_formbrack = proto->lex_formbrack;
10434 parser->lex_inpat = proto->lex_inpat;
10435 parser->lex_inwhat = proto->lex_inwhat;
10436 parser->lex_op = proto->lex_op;
10437 parser->lex_repl = sv_dup_inc(proto->lex_repl, param);
10438 parser->lex_starts = proto->lex_starts;
10439 parser->lex_stuff = sv_dup_inc(proto->lex_stuff, param);
10440 parser->multi_close = proto->multi_close;
10441 parser->multi_open = proto->multi_open;
10442 parser->multi_start = proto->multi_start;
670a9cb2 10443 parser->multi_end = proto->multi_end;
199e78b7
DM
10444 parser->pending_ident = proto->pending_ident;
10445 parser->preambled = proto->preambled;
10446 parser->sublex_info = proto->sublex_info; /* XXX not quite right */
bdc0bf6f 10447 parser->linestr = sv_dup_inc(proto->linestr, param);
53a7735b
DM
10448 parser->expect = proto->expect;
10449 parser->copline = proto->copline;
f06b5848 10450 parser->last_lop_op = proto->last_lop_op;
bc177e6b 10451 parser->lex_state = proto->lex_state;
2f9285f8 10452 parser->rsfp = fp_dup(proto->rsfp, '<', param);
5486870f
DM
10453 /* rsfp_filters entries have fake IoDIRP() */
10454 parser->rsfp_filters= av_dup_inc(proto->rsfp_filters, param);
12bd6ede
DM
10455 parser->in_my = proto->in_my;
10456 parser->in_my_stash = hv_dup(proto->in_my_stash, param);
13765c85 10457 parser->error_count = proto->error_count;
bc177e6b 10458
53a7735b 10459
f06b5848
DM
10460 parser->linestr = sv_dup_inc(proto->linestr, param);
10461
10462 {
1e05feb3
AL
10463 char * const ols = SvPVX(proto->linestr);
10464 char * const ls = SvPVX(parser->linestr);
f06b5848
DM
10465
10466 parser->bufptr = ls + (proto->bufptr >= ols ?
10467 proto->bufptr - ols : 0);
10468 parser->oldbufptr = ls + (proto->oldbufptr >= ols ?
10469 proto->oldbufptr - ols : 0);
10470 parser->oldoldbufptr= ls + (proto->oldoldbufptr >= ols ?
10471 proto->oldoldbufptr - ols : 0);
10472 parser->linestart = ls + (proto->linestart >= ols ?
10473 proto->linestart - ols : 0);
10474 parser->last_uni = ls + (proto->last_uni >= ols ?
10475 proto->last_uni - ols : 0);
10476 parser->last_lop = ls + (proto->last_lop >= ols ?
10477 proto->last_lop - ols : 0);
10478
10479 parser->bufend = ls + SvCUR(parser->linestr);
10480 }
199e78b7 10481
14047fc9
DM
10482 Copy(proto->tokenbuf, parser->tokenbuf, 256, char);
10483
2f9285f8 10484
199e78b7
DM
10485#ifdef PERL_MAD
10486 parser->endwhite = proto->endwhite;
10487 parser->faketokens = proto->faketokens;
10488 parser->lasttoke = proto->lasttoke;
10489 parser->nextwhite = proto->nextwhite;
10490 parser->realtokenstart = proto->realtokenstart;
10491 parser->skipwhite = proto->skipwhite;
10492 parser->thisclose = proto->thisclose;
10493 parser->thismad = proto->thismad;
10494 parser->thisopen = proto->thisopen;
10495 parser->thisstuff = proto->thisstuff;
10496 parser->thistoken = proto->thistoken;
10497 parser->thiswhite = proto->thiswhite;
fb205e7a
DM
10498
10499 Copy(proto->nexttoke, parser->nexttoke, 5, NEXTTOKE);
10500 parser->curforce = proto->curforce;
10501#else
10502 Copy(proto->nextval, parser->nextval, 5, YYSTYPE);
10503 Copy(proto->nexttype, parser->nexttype, 5, I32);
10504 parser->nexttoke = proto->nexttoke;
199e78b7 10505#endif
f0c5aa00
DM
10506
10507 /* XXX should clone saved_curcop here, but we aren't passed
10508 * proto_perl; so do it in perl_clone_using instead */
10509
199e78b7
DM
10510 return parser;
10511}
10512
d2d73c3e 10513
d2d73c3e 10514/* duplicate a file handle */
645c22ef 10515
1d7c1841 10516PerlIO *
3be3cdd6 10517Perl_fp_dup(pTHX_ PerlIO *const fp, const char type, CLONE_PARAMS *const param)
1d7c1841
GS
10518{
10519 PerlIO *ret;
53c1dcc0 10520
7918f24d 10521 PERL_ARGS_ASSERT_FP_DUP;
53c1dcc0 10522 PERL_UNUSED_ARG(type);
73d840c0 10523
1d7c1841
GS
10524 if (!fp)
10525 return (PerlIO*)NULL;
10526
10527 /* look for it in the table first */
10528 ret = (PerlIO*)ptr_table_fetch(PL_ptr_table, fp);
10529 if (ret)
10530 return ret;
10531
10532 /* create anew and remember what it is */
ecdeb87c 10533 ret = PerlIO_fdupopen(aTHX_ fp, param, PERLIO_DUP_CLONE);
1d7c1841
GS
10534 ptr_table_store(PL_ptr_table, fp, ret);
10535 return ret;
10536}
10537
645c22ef
DM
10538/* duplicate a directory handle */
10539
1d7c1841 10540DIR *
66ceb532 10541Perl_dirp_dup(pTHX_ DIR *const dp)
1d7c1841 10542{
96a5add6 10543 PERL_UNUSED_CONTEXT;
1d7c1841
GS
10544 if (!dp)
10545 return (DIR*)NULL;
10546 /* XXX TODO */
10547 return dp;
10548}
10549
ff276b08 10550/* duplicate a typeglob */
645c22ef 10551
1d7c1841 10552GP *
66ceb532 10553Perl_gp_dup(pTHX_ GP *const gp, CLONE_PARAMS *const param)
1d7c1841
GS
10554{
10555 GP *ret;
b37c2d43 10556
7918f24d
NC
10557 PERL_ARGS_ASSERT_GP_DUP;
10558
1d7c1841
GS
10559 if (!gp)
10560 return (GP*)NULL;
10561 /* look for it in the table first */
10562 ret = (GP*)ptr_table_fetch(PL_ptr_table, gp);
10563 if (ret)
10564 return ret;
10565
10566 /* create anew and remember what it is */
a02a5408 10567 Newxz(ret, 1, GP);
1d7c1841
GS
10568 ptr_table_store(PL_ptr_table, gp, ret);
10569
10570 /* clone */
46d65037
NC
10571 /* ret->gp_refcnt must be 0 before any other dups are called. We're relying
10572 on Newxz() to do this for us. */
d2d73c3e
AB
10573 ret->gp_sv = sv_dup_inc(gp->gp_sv, param);
10574 ret->gp_io = io_dup_inc(gp->gp_io, param);
10575 ret->gp_form = cv_dup_inc(gp->gp_form, param);
10576 ret->gp_av = av_dup_inc(gp->gp_av, param);
10577 ret->gp_hv = hv_dup_inc(gp->gp_hv, param);
10578 ret->gp_egv = gv_dup(gp->gp_egv, param);/* GvEGV is not refcounted */
10579 ret->gp_cv = cv_dup_inc(gp->gp_cv, param);
1d7c1841 10580 ret->gp_cvgen = gp->gp_cvgen;
1d7c1841 10581 ret->gp_line = gp->gp_line;
566771cc 10582 ret->gp_file_hek = hek_dup(gp->gp_file_hek, param);
1d7c1841
GS
10583 return ret;
10584}
10585
645c22ef
DM
10586/* duplicate a chain of magic */
10587
1d7c1841 10588MAGIC *
b88ec9b8 10589Perl_mg_dup(pTHX_ MAGIC *mg, CLONE_PARAMS *const param)
1d7c1841 10590{
c160a186 10591 MAGIC *mgret = NULL;
0228edf6 10592 MAGIC **mgprev_p = &mgret;
7918f24d
NC
10593
10594 PERL_ARGS_ASSERT_MG_DUP;
10595
1d7c1841
GS
10596 for (; mg; mg = mg->mg_moremagic) {
10597 MAGIC *nmg;
45f7fcc8 10598 Newx(nmg, 1, MAGIC);
0228edf6
NC
10599 *mgprev_p = nmg;
10600 mgprev_p = &(nmg->mg_moremagic);
10601
45f7fcc8
NC
10602 /* There was a comment "XXX copy dynamic vtable?" but as we don't have
10603 dynamic vtables, I'm not sure why Sarathy wrote it. The comment dates
10604 from the original commit adding Perl_mg_dup() - revision 4538.
10605 Similarly there is the annotation "XXX random ptr?" next to the
10606 assignment to nmg->mg_ptr. */
10607 *nmg = *mg;
10608
288b8c02 10609 /* FIXME for plugins
45f7fcc8
NC
10610 if (nmg->mg_type == PERL_MAGIC_qr) {
10611 nmg->mg_obj = MUTABLE_SV(CALLREGDUPE((REGEXP*)nmg->mg_obj, param));
1d7c1841 10612 }
288b8c02
NC
10613 else
10614 */
45f7fcc8 10615 if(nmg->mg_type == PERL_MAGIC_backref) {
d7cbc7b5
NC
10616 /* The backref AV has its reference count deliberately bumped by
10617 1. */
502c6561 10618 nmg->mg_obj
45f7fcc8 10619 = SvREFCNT_inc(av_dup_inc((const AV *) nmg->mg_obj, param));
05bd4103 10620 }
1d7c1841 10621 else {
45f7fcc8
NC
10622 nmg->mg_obj = (nmg->mg_flags & MGf_REFCOUNTED)
10623 ? sv_dup_inc(nmg->mg_obj, param)
10624 : sv_dup(nmg->mg_obj, param);
10625 }
10626
10627 if (nmg->mg_ptr && nmg->mg_type != PERL_MAGIC_regex_global) {
10628 if (nmg->mg_len > 0) {
10629 nmg->mg_ptr = SAVEPVN(nmg->mg_ptr, nmg->mg_len);
10630 if (nmg->mg_type == PERL_MAGIC_overload_table &&
10631 AMT_AMAGIC((AMT*)nmg->mg_ptr))
14befaf4 10632 {
0bcc34c2 10633 AMT * const namtp = (AMT*)nmg->mg_ptr;
538f2e76
NC
10634 sv_dup_inc_multiple((SV**)(namtp->table),
10635 (SV**)(namtp->table), NofAMmeth, param);
1d7c1841
GS
10636 }
10637 }
45f7fcc8
NC
10638 else if (nmg->mg_len == HEf_SVKEY)
10639 nmg->mg_ptr = (char*)sv_dup_inc((const SV *)nmg->mg_ptr, param);
1d7c1841 10640 }
45f7fcc8 10641 if ((nmg->mg_flags & MGf_DUP) && nmg->mg_virtual && nmg->mg_virtual->svt_dup) {
68795e93
NIS
10642 CALL_FPTR(nmg->mg_virtual->svt_dup)(aTHX_ nmg, param);
10643 }
1d7c1841
GS
10644 }
10645 return mgret;
10646}
10647
4674ade5
NC
10648#endif /* USE_ITHREADS */
10649
645c22ef
DM
10650/* create a new pointer-mapping table */
10651
1d7c1841
GS
10652PTR_TBL_t *
10653Perl_ptr_table_new(pTHX)
10654{
10655 PTR_TBL_t *tbl;
96a5add6
AL
10656 PERL_UNUSED_CONTEXT;
10657
b3a120bf 10658 Newx(tbl, 1, PTR_TBL_t);
1d7c1841
GS
10659 tbl->tbl_max = 511;
10660 tbl->tbl_items = 0;
a02a5408 10661 Newxz(tbl->tbl_ary, tbl->tbl_max + 1, PTR_TBL_ENT_t*);
1d7c1841
GS
10662 return tbl;
10663}
10664
7119fd33
NC
10665#define PTR_TABLE_HASH(ptr) \
10666 ((PTR2UV(ptr) >> 3) ^ (PTR2UV(ptr) >> (3 + 7)) ^ (PTR2UV(ptr) >> (3 + 17)))
134ca3d6 10667
93e68bfb
JC
10668/*
10669 we use the PTE_SVSLOT 'reservation' made above, both here (in the
10670 following define) and at call to new_body_inline made below in
10671 Perl_ptr_table_store()
10672 */
10673
10674#define del_pte(p) del_body_type(p, PTE_SVSLOT)
32e691d0 10675
645c22ef
DM
10676/* map an existing pointer using a table */
10677
7bf61b54 10678STATIC PTR_TBL_ENT_t *
1eb6e4ca 10679S_ptr_table_find(PTR_TBL_t *const tbl, const void *const sv)
7918f24d 10680{
1d7c1841 10681 PTR_TBL_ENT_t *tblent;
4373e329 10682 const UV hash = PTR_TABLE_HASH(sv);
7918f24d
NC
10683
10684 PERL_ARGS_ASSERT_PTR_TABLE_FIND;
10685
1d7c1841
GS
10686 tblent = tbl->tbl_ary[hash & tbl->tbl_max];
10687 for (; tblent; tblent = tblent->next) {
10688 if (tblent->oldval == sv)
7bf61b54 10689 return tblent;
1d7c1841 10690 }
d4c19fe8 10691 return NULL;
7bf61b54
NC
10692}
10693
10694void *
1eb6e4ca 10695Perl_ptr_table_fetch(pTHX_ PTR_TBL_t *const tbl, const void *const sv)
7bf61b54 10696{
b0e6ae5b 10697 PTR_TBL_ENT_t const *const tblent = ptr_table_find(tbl, sv);
7918f24d
NC
10698
10699 PERL_ARGS_ASSERT_PTR_TABLE_FETCH;
96a5add6 10700 PERL_UNUSED_CONTEXT;
7918f24d 10701
d4c19fe8 10702 return tblent ? tblent->newval : NULL;
1d7c1841
GS
10703}
10704
645c22ef
DM
10705/* add a new entry to a pointer-mapping table */
10706
1d7c1841 10707void
1eb6e4ca 10708Perl_ptr_table_store(pTHX_ PTR_TBL_t *const tbl, const void *const oldsv, void *const newsv)
1d7c1841 10709{
0c9fdfe0 10710 PTR_TBL_ENT_t *tblent = ptr_table_find(tbl, oldsv);
7918f24d
NC
10711
10712 PERL_ARGS_ASSERT_PTR_TABLE_STORE;
96a5add6 10713 PERL_UNUSED_CONTEXT;
1d7c1841 10714
7bf61b54
NC
10715 if (tblent) {
10716 tblent->newval = newsv;
10717 } else {
10718 const UV entry = PTR_TABLE_HASH(oldsv) & tbl->tbl_max;
10719
d2a0f284
JC
10720 new_body_inline(tblent, PTE_SVSLOT);
10721
7bf61b54
NC
10722 tblent->oldval = oldsv;
10723 tblent->newval = newsv;
10724 tblent->next = tbl->tbl_ary[entry];
10725 tbl->tbl_ary[entry] = tblent;
10726 tbl->tbl_items++;
10727 if (tblent->next && tbl->tbl_items > tbl->tbl_max)
10728 ptr_table_split(tbl);
1d7c1841 10729 }
1d7c1841
GS
10730}
10731
645c22ef
DM
10732/* double the hash bucket size of an existing ptr table */
10733
1d7c1841 10734void
1eb6e4ca 10735Perl_ptr_table_split(pTHX_ PTR_TBL_t *const tbl)
1d7c1841
GS
10736{
10737 PTR_TBL_ENT_t **ary = tbl->tbl_ary;
4373e329 10738 const UV oldsize = tbl->tbl_max + 1;
1d7c1841
GS
10739 UV newsize = oldsize * 2;
10740 UV i;
7918f24d
NC
10741
10742 PERL_ARGS_ASSERT_PTR_TABLE_SPLIT;
96a5add6 10743 PERL_UNUSED_CONTEXT;
1d7c1841
GS
10744
10745 Renew(ary, newsize, PTR_TBL_ENT_t*);
10746 Zero(&ary[oldsize], newsize-oldsize, PTR_TBL_ENT_t*);
10747 tbl->tbl_max = --newsize;
10748 tbl->tbl_ary = ary;
10749 for (i=0; i < oldsize; i++, ary++) {
10750 PTR_TBL_ENT_t **curentp, **entp, *ent;
10751 if (!*ary)
10752 continue;
10753 curentp = ary + oldsize;
10754 for (entp = ary, ent = *ary; ent; ent = *entp) {
134ca3d6 10755 if ((newsize & PTR_TABLE_HASH(ent->oldval)) != i) {
1d7c1841
GS
10756 *entp = ent->next;
10757 ent->next = *curentp;
10758 *curentp = ent;
10759 continue;
10760 }
10761 else
10762 entp = &ent->next;
10763 }
10764 }
10765}
10766
645c22ef
DM
10767/* remove all the entries from a ptr table */
10768
a0739874 10769void
1eb6e4ca 10770Perl_ptr_table_clear(pTHX_ PTR_TBL_t *const tbl)
a0739874 10771{
d5cefff9 10772 if (tbl && tbl->tbl_items) {
c445ea15 10773 register PTR_TBL_ENT_t * const * const array = tbl->tbl_ary;
d5cefff9 10774 UV riter = tbl->tbl_max;
a0739874 10775
d5cefff9
NC
10776 do {
10777 PTR_TBL_ENT_t *entry = array[riter];
ab1e7f95 10778
d5cefff9 10779 while (entry) {
00b6aa41 10780 PTR_TBL_ENT_t * const oentry = entry;
d5cefff9
NC
10781 entry = entry->next;
10782 del_pte(oentry);
10783 }
10784 } while (riter--);
a0739874 10785
d5cefff9
NC
10786 tbl->tbl_items = 0;
10787 }
a0739874
DM
10788}
10789
645c22ef
DM
10790/* clear and free a ptr table */
10791
a0739874 10792void
1eb6e4ca 10793Perl_ptr_table_free(pTHX_ PTR_TBL_t *const tbl)
a0739874
DM
10794{
10795 if (!tbl) {
10796 return;
10797 }
10798 ptr_table_clear(tbl);
10799 Safefree(tbl->tbl_ary);
10800 Safefree(tbl);
10801}
10802
4674ade5 10803#if defined(USE_ITHREADS)
5bd07a3d 10804
83841fad 10805void
1eb6e4ca 10806Perl_rvpv_dup(pTHX_ SV *const dstr, const SV *const sstr, CLONE_PARAMS *const param)
83841fad 10807{
7918f24d
NC
10808 PERL_ARGS_ASSERT_RVPV_DUP;
10809
83841fad 10810 if (SvROK(sstr)) {
b162af07 10811 SvRV_set(dstr, SvWEAKREF(sstr)
f19a12a3
MHM
10812 ? sv_dup(SvRV_const(sstr), param)
10813 : sv_dup_inc(SvRV_const(sstr), param));
f880fe2f 10814
83841fad 10815 }
3f7c398e 10816 else if (SvPVX_const(sstr)) {
83841fad
NIS
10817 /* Has something there */
10818 if (SvLEN(sstr)) {
68795e93 10819 /* Normal PV - clone whole allocated space */
3f7c398e 10820 SvPV_set(dstr, SAVEPVN(SvPVX_const(sstr), SvLEN(sstr)-1));
d3d0e6f1
NC
10821 if (SvREADONLY(sstr) && SvFAKE(sstr)) {
10822 /* Not that normal - actually sstr is copy on write.
10823 But we are a true, independant SV, so: */
10824 SvREADONLY_off(dstr);
10825 SvFAKE_off(dstr);
10826 }
68795e93 10827 }
83841fad
NIS
10828 else {
10829 /* Special case - not normally malloced for some reason */
f7877b28
NC
10830 if (isGV_with_GP(sstr)) {
10831 /* Don't need to do anything here. */
10832 }
10833 else if ((SvREADONLY(sstr) && SvFAKE(sstr))) {
ef10be65
NC
10834 /* A "shared" PV - clone it as "shared" PV */
10835 SvPV_set(dstr,
10836 HEK_KEY(hek_dup(SvSHARED_HEK_FROM_PV(SvPVX_const(sstr)),
10837 param)));
83841fad
NIS
10838 }
10839 else {
10840 /* Some other special case - random pointer */
d2c6dc5e 10841 SvPV_set(dstr, (char *) SvPVX_const(sstr));
d3d0e6f1 10842 }
83841fad
NIS
10843 }
10844 }
10845 else {
4608196e 10846 /* Copy the NULL */
4df7f6af 10847 SvPV_set(dstr, NULL);
83841fad
NIS
10848 }
10849}
10850
538f2e76
NC
10851/* duplicate a list of SVs. source and dest may point to the same memory. */
10852static SV **
10853S_sv_dup_inc_multiple(pTHX_ SV *const *source, SV **dest,
10854 SSize_t items, CLONE_PARAMS *const param)
10855{
10856 PERL_ARGS_ASSERT_SV_DUP_INC_MULTIPLE;
10857
10858 while (items-- > 0) {
10859 *dest++ = sv_dup_inc(*source++, param);
10860 }
10861
10862 return dest;
10863}
10864
662fb8b2
NC
10865/* duplicate an SV of any type (including AV, HV etc) */
10866
1d7c1841 10867SV *
1eb6e4ca 10868Perl_sv_dup(pTHX_ const SV *const sstr, CLONE_PARAMS *const param)
1d7c1841 10869{
27da23d5 10870 dVAR;
1d7c1841
GS
10871 SV *dstr;
10872
7918f24d
NC
10873 PERL_ARGS_ASSERT_SV_DUP;
10874
bfd95973
NC
10875 if (!sstr)
10876 return NULL;
10877 if (SvTYPE(sstr) == SVTYPEMASK) {
10878#ifdef DEBUG_LEAKING_SCALARS_ABORT
10879 abort();
10880#endif
6136c704 10881 return NULL;
bfd95973 10882 }
1d7c1841 10883 /* look for it in the table first */
daba3364 10884 dstr = MUTABLE_SV(ptr_table_fetch(PL_ptr_table, sstr));
1d7c1841
GS
10885 if (dstr)
10886 return dstr;
10887
0405e91e
AB
10888 if(param->flags & CLONEf_JOIN_IN) {
10889 /** We are joining here so we don't want do clone
10890 something that is bad **/
eb86f8b3 10891 if (SvTYPE(sstr) == SVt_PVHV) {
9bde8eb0 10892 const HEK * const hvname = HvNAME_HEK(sstr);
eb86f8b3
AL
10893 if (hvname)
10894 /** don't clone stashes if they already exist **/
daba3364 10895 return MUTABLE_SV(gv_stashpvn(HEK_KEY(hvname), HEK_LEN(hvname), 0));
0405e91e
AB
10896 }
10897 }
10898
1d7c1841
GS
10899 /* create anew and remember what it is */
10900 new_SV(dstr);
fd0854ff
DM
10901
10902#ifdef DEBUG_LEAKING_SCALARS
10903 dstr->sv_debug_optype = sstr->sv_debug_optype;
10904 dstr->sv_debug_line = sstr->sv_debug_line;
10905 dstr->sv_debug_inpad = sstr->sv_debug_inpad;
10906 dstr->sv_debug_cloned = 1;
fd0854ff 10907 dstr->sv_debug_file = savepv(sstr->sv_debug_file);
fd0854ff
DM
10908#endif
10909
1d7c1841
GS
10910 ptr_table_store(PL_ptr_table, sstr, dstr);
10911
10912 /* clone */
10913 SvFLAGS(dstr) = SvFLAGS(sstr);
10914 SvFLAGS(dstr) &= ~SVf_OOK; /* don't propagate OOK hack */
10915 SvREFCNT(dstr) = 0; /* must be before any other dups! */
10916
10917#ifdef DEBUGGING
3f7c398e 10918 if (SvANY(sstr) && PL_watch_pvx && SvPVX_const(sstr) == PL_watch_pvx)
1d7c1841 10919 PerlIO_printf(Perl_debug_log, "watch at %p hit, found string \"%s\"\n",
6c9570dc 10920 (void*)PL_watch_pvx, SvPVX_const(sstr));
1d7c1841
GS
10921#endif
10922
9660f481
DM
10923 /* don't clone objects whose class has asked us not to */
10924 if (SvOBJECT(sstr) && ! (SvFLAGS(SvSTASH(sstr)) & SVphv_CLONEABLE)) {
33de8e4a 10925 SvFLAGS(dstr) = 0;
9660f481
DM
10926 return dstr;
10927 }
10928
1d7c1841
GS
10929 switch (SvTYPE(sstr)) {
10930 case SVt_NULL:
10931 SvANY(dstr) = NULL;
10932 break;
10933 case SVt_IV:
339049b0 10934 SvANY(dstr) = (XPVIV*)((char*)&(dstr->sv_u.svu_iv) - STRUCT_OFFSET(XPVIV, xiv_iv));
4df7f6af
NC
10935 if(SvROK(sstr)) {
10936 Perl_rvpv_dup(aTHX_ dstr, sstr, param);
10937 } else {
10938 SvIV_set(dstr, SvIVX(sstr));
10939 }
1d7c1841
GS
10940 break;
10941 case SVt_NV:
10942 SvANY(dstr) = new_XNV();
9d6ce603 10943 SvNV_set(dstr, SvNVX(sstr));
1d7c1841 10944 break;
cecf5685 10945 /* case SVt_BIND: */
662fb8b2
NC
10946 default:
10947 {
10948 /* These are all the types that need complex bodies allocating. */
662fb8b2 10949 void *new_body;
2bcc16b3
NC
10950 const svtype sv_type = SvTYPE(sstr);
10951 const struct body_details *const sv_type_details
10952 = bodies_by_type + sv_type;
662fb8b2 10953
93e68bfb 10954 switch (sv_type) {
662fb8b2 10955 default:
bb263b4e 10956 Perl_croak(aTHX_ "Bizarre SvTYPE [%" IVdf "]", (IV)SvTYPE(sstr));
662fb8b2
NC
10957 break;
10958
662fb8b2 10959 case SVt_PVGV:
c22188b4
NC
10960 case SVt_PVIO:
10961 case SVt_PVFM:
10962 case SVt_PVHV:
10963 case SVt_PVAV:
662fb8b2 10964 case SVt_PVCV:
662fb8b2 10965 case SVt_PVLV:
5c35adbb 10966 case SVt_REGEXP:
662fb8b2 10967 case SVt_PVMG:
662fb8b2 10968 case SVt_PVNV:
662fb8b2 10969 case SVt_PVIV:
662fb8b2 10970 case SVt_PV:
d2a0f284 10971 assert(sv_type_details->body_size);
c22188b4 10972 if (sv_type_details->arena) {
d2a0f284 10973 new_body_inline(new_body, sv_type);
c22188b4 10974 new_body
b9502f15 10975 = (void*)((char*)new_body - sv_type_details->offset);
c22188b4
NC
10976 } else {
10977 new_body = new_NOARENA(sv_type_details);
10978 }
1d7c1841 10979 }
662fb8b2
NC
10980 assert(new_body);
10981 SvANY(dstr) = new_body;
10982
2bcc16b3 10983#ifndef PURIFY
b9502f15
NC
10984 Copy(((char*)SvANY(sstr)) + sv_type_details->offset,
10985 ((char*)SvANY(dstr)) + sv_type_details->offset,
f32993d6 10986 sv_type_details->copy, char);
2bcc16b3
NC
10987#else
10988 Copy(((char*)SvANY(sstr)),
10989 ((char*)SvANY(dstr)),
d2a0f284 10990 sv_type_details->body_size + sv_type_details->offset, char);
2bcc16b3 10991#endif
662fb8b2 10992
f7877b28
NC
10993 if (sv_type != SVt_PVAV && sv_type != SVt_PVHV
10994 && !isGV_with_GP(dstr))
662fb8b2
NC
10995 Perl_rvpv_dup(aTHX_ dstr, sstr, param);
10996
10997 /* The Copy above means that all the source (unduplicated) pointers
10998 are now in the destination. We can check the flags and the
10999 pointers in either, but it's possible that there's less cache
11000 missing by always going for the destination.
11001 FIXME - instrument and check that assumption */
f32993d6 11002 if (sv_type >= SVt_PVMG) {
885ffcb3 11003 if ((sv_type == SVt_PVMG) && SvPAD_OUR(dstr)) {
73d95100 11004 SvOURSTASH_set(dstr, hv_dup_inc(SvOURSTASH(dstr), param));
e736a858 11005 } else if (SvMAGIC(dstr))
662fb8b2
NC
11006 SvMAGIC_set(dstr, mg_dup(SvMAGIC(dstr), param));
11007 if (SvSTASH(dstr))
11008 SvSTASH_set(dstr, hv_dup_inc(SvSTASH(dstr), param));
1d7c1841 11009 }
662fb8b2 11010
f32993d6
NC
11011 /* The cast silences a GCC warning about unhandled types. */
11012 switch ((int)sv_type) {
662fb8b2
NC
11013 case SVt_PV:
11014 break;
11015 case SVt_PVIV:
11016 break;
11017 case SVt_PVNV:
11018 break;
11019 case SVt_PVMG:
11020 break;
5c35adbb 11021 case SVt_REGEXP:
288b8c02 11022 /* FIXME for plugins */
d2f13c59 11023 re_dup_guts((REGEXP*) sstr, (REGEXP*) dstr, param);
f708cfc1 11024 break;
662fb8b2
NC
11025 case SVt_PVLV:
11026 /* XXX LvTARGOFF sometimes holds PMOP* when DEBUGGING */
11027 if (LvTYPE(dstr) == 't') /* for tie: unrefcnted fake (SV**) */
11028 LvTARG(dstr) = dstr;
11029 else if (LvTYPE(dstr) == 'T') /* for tie: fake HE */
daba3364 11030 LvTARG(dstr) = MUTABLE_SV(he_dup((HE*)LvTARG(dstr), 0, param));
662fb8b2
NC
11031 else
11032 LvTARG(dstr) = sv_dup_inc(LvTARG(dstr), param);
662fb8b2 11033 case SVt_PVGV:
cecf5685 11034 if(isGV_with_GP(sstr)) {
566771cc 11035 GvNAME_HEK(dstr) = hek_dup(GvNAME_HEK(dstr), param);
39cb70dc
NC
11036 /* Don't call sv_add_backref here as it's going to be
11037 created as part of the magic cloning of the symbol
11038 table. */
f7877b28
NC
11039 /* Danger Will Robinson - GvGP(dstr) isn't initialised
11040 at the point of this comment. */
39cb70dc 11041 GvSTASH(dstr) = hv_dup(GvSTASH(dstr), param);
f7877b28
NC
11042 GvGP(dstr) = gp_dup(GvGP(sstr), param);
11043 (void)GpREFCNT_inc(GvGP(dstr));
11044 } else
11045 Perl_rvpv_dup(aTHX_ dstr, sstr, param);
662fb8b2
NC
11046 break;
11047 case SVt_PVIO:
11048 IoIFP(dstr) = fp_dup(IoIFP(dstr), IoTYPE(dstr), param);
11049 if (IoOFP(dstr) == IoIFP(sstr))
11050 IoOFP(dstr) = IoIFP(dstr);
11051 else
11052 IoOFP(dstr) = fp_dup(IoOFP(dstr), IoTYPE(dstr), param);
5486870f 11053 /* PL_parser->rsfp_filters entries have fake IoDIRP() */
662fb8b2
NC
11054 if(IoFLAGS(dstr) & IOf_FAKE_DIRP) {
11055 /* I have no idea why fake dirp (rsfps)
11056 should be treated differently but otherwise
11057 we end up with leaks -- sky*/
11058 IoTOP_GV(dstr) = gv_dup_inc(IoTOP_GV(dstr), param);
11059 IoFMT_GV(dstr) = gv_dup_inc(IoFMT_GV(dstr), param);
11060 IoBOTTOM_GV(dstr) = gv_dup_inc(IoBOTTOM_GV(dstr), param);
11061 } else {
11062 IoTOP_GV(dstr) = gv_dup(IoTOP_GV(dstr), param);
11063 IoFMT_GV(dstr) = gv_dup(IoFMT_GV(dstr), param);
11064 IoBOTTOM_GV(dstr) = gv_dup(IoBOTTOM_GV(dstr), param);
100ce7e1
NC
11065 if (IoDIRP(dstr)) {
11066 IoDIRP(dstr) = dirp_dup(IoDIRP(dstr));
11067 } else {
6f207bd3 11068 NOOP;
100ce7e1
NC
11069 /* IoDIRP(dstr) is already a copy of IoDIRP(sstr) */
11070 }
662fb8b2
NC
11071 }
11072 IoTOP_NAME(dstr) = SAVEPV(IoTOP_NAME(dstr));
11073 IoFMT_NAME(dstr) = SAVEPV(IoFMT_NAME(dstr));
11074 IoBOTTOM_NAME(dstr) = SAVEPV(IoBOTTOM_NAME(dstr));
11075 break;
11076 case SVt_PVAV:
2779b694
KB
11077 /* avoid cloning an empty array */
11078 if (AvARRAY((const AV *)sstr) && AvFILLp((const AV *)sstr) >= 0) {
662fb8b2 11079 SV **dst_ary, **src_ary;
502c6561 11080 SSize_t items = AvFILLp((const AV *)sstr) + 1;
662fb8b2 11081
502c6561
NC
11082 src_ary = AvARRAY((const AV *)sstr);
11083 Newxz(dst_ary, AvMAX((const AV *)sstr)+1, SV*);
662fb8b2 11084 ptr_table_store(PL_ptr_table, src_ary, dst_ary);
502c6561
NC
11085 AvARRAY(MUTABLE_AV(dstr)) = dst_ary;
11086 AvALLOC((const AV *)dstr) = dst_ary;
11087 if (AvREAL((const AV *)sstr)) {
538f2e76
NC
11088 dst_ary = sv_dup_inc_multiple(src_ary, dst_ary, items,
11089 param);
662fb8b2
NC
11090 }
11091 else {
11092 while (items-- > 0)
11093 *dst_ary++ = sv_dup(*src_ary++, param);
11094 }
502c6561 11095 items = AvMAX((const AV *)sstr) - AvFILLp((const AV *)sstr);
662fb8b2
NC
11096 while (items-- > 0) {
11097 *dst_ary++ = &PL_sv_undef;
11098 }
bfcb3514 11099 }
662fb8b2 11100 else {
502c6561
NC
11101 AvARRAY(MUTABLE_AV(dstr)) = NULL;
11102 AvALLOC((const AV *)dstr) = (SV**)NULL;
2779b694
KB
11103 AvMAX( (const AV *)dstr) = -1;
11104 AvFILLp((const AV *)dstr) = -1;
b79f7545 11105 }
662fb8b2
NC
11106 break;
11107 case SVt_PVHV:
1d193675 11108 if (HvARRAY((const HV *)sstr)) {
7e265ef3
AL
11109 STRLEN i = 0;
11110 const bool sharekeys = !!HvSHAREKEYS(sstr);
11111 XPVHV * const dxhv = (XPVHV*)SvANY(dstr);
11112 XPVHV * const sxhv = (XPVHV*)SvANY(sstr);
11113 char *darray;
11114 Newx(darray, PERL_HV_ARRAY_ALLOC_BYTES(dxhv->xhv_max+1)
11115 + (SvOOK(sstr) ? sizeof(struct xpvhv_aux) : 0),
11116 char);
11117 HvARRAY(dstr) = (HE**)darray;
11118 while (i <= sxhv->xhv_max) {
11119 const HE * const source = HvARRAY(sstr)[i];
11120 HvARRAY(dstr)[i] = source
11121 ? he_dup(source, sharekeys, param) : 0;
11122 ++i;
11123 }
11124 if (SvOOK(sstr)) {
11125 HEK *hvname;
11126 const struct xpvhv_aux * const saux = HvAUX(sstr);
11127 struct xpvhv_aux * const daux = HvAUX(dstr);
11128 /* This flag isn't copied. */
11129 /* SvOOK_on(hv) attacks the IV flags. */
11130 SvFLAGS(dstr) |= SVf_OOK;
11131
11132 hvname = saux->xhv_name;
566771cc 11133 daux->xhv_name = hek_dup(hvname, param);
7e265ef3
AL
11134
11135 daux->xhv_riter = saux->xhv_riter;
11136 daux->xhv_eiter = saux->xhv_eiter
11137 ? he_dup(saux->xhv_eiter,
11138 (bool)!!HvSHAREKEYS(sstr), param) : 0;
b17f5ab7 11139 /* backref array needs refcnt=2; see sv_add_backref */
7e265ef3
AL
11140 daux->xhv_backreferences =
11141 saux->xhv_backreferences
502c6561 11142 ? MUTABLE_AV(SvREFCNT_inc(
daba3364 11143 sv_dup_inc((const SV *)saux->xhv_backreferences, param)))
86f55936 11144 : 0;
e1a479c5
BB
11145
11146 daux->xhv_mro_meta = saux->xhv_mro_meta
11147 ? mro_meta_dup(saux->xhv_mro_meta, param)
11148 : 0;
11149
7e265ef3
AL
11150 /* Record stashes for possible cloning in Perl_clone(). */
11151 if (hvname)
11152 av_push(param->stashes, dstr);
662fb8b2 11153 }
662fb8b2 11154 }
7e265ef3 11155 else
85fbaab2 11156 HvARRAY(MUTABLE_HV(dstr)) = NULL;
662fb8b2 11157 break;
662fb8b2 11158 case SVt_PVCV:
bb172083
NC
11159 if (!(param->flags & CLONEf_COPY_STACKS)) {
11160 CvDEPTH(dstr) = 0;
11161 }
11162 case SVt_PVFM:
662fb8b2
NC
11163 /* NOTE: not refcounted */
11164 CvSTASH(dstr) = hv_dup(CvSTASH(dstr), param);
11165 OP_REFCNT_LOCK;
d04ba589
NC
11166 if (!CvISXSUB(dstr))
11167 CvROOT(dstr) = OpREFCNT_inc(CvROOT(dstr));
662fb8b2 11168 OP_REFCNT_UNLOCK;
cfae286e 11169 if (CvCONST(dstr) && CvISXSUB(dstr)) {
d32faaf3 11170 CvXSUBANY(dstr).any_ptr =
daba3364 11171 sv_dup_inc((const SV *)CvXSUBANY(dstr).any_ptr, param);
662fb8b2
NC
11172 }
11173 /* don't dup if copying back - CvGV isn't refcounted, so the
11174 * duped GV may never be freed. A bit of a hack! DAPM */
11175 CvGV(dstr) = (param->flags & CLONEf_JOIN_IN) ?
a0714e2c 11176 NULL : gv_dup(CvGV(dstr), param) ;
662fb8b2
NC
11177 PAD_DUP(CvPADLIST(dstr), CvPADLIST(sstr), param);
11178 CvOUTSIDE(dstr) =
11179 CvWEAKOUTSIDE(sstr)
11180 ? cv_dup( CvOUTSIDE(dstr), param)
11181 : cv_dup_inc(CvOUTSIDE(dstr), param);
aed2304a 11182 if (!CvISXSUB(dstr))
662fb8b2
NC
11183 CvFILE(dstr) = SAVEPV(CvFILE(dstr));
11184 break;
bfcb3514 11185 }
1d7c1841 11186 }
1d7c1841
GS
11187 }
11188
11189 if (SvOBJECT(dstr) && SvTYPE(dstr) != SVt_PVIO)
11190 ++PL_sv_objcount;
11191
11192 return dstr;
d2d73c3e 11193 }
1d7c1841 11194
645c22ef
DM
11195/* duplicate a context */
11196
1d7c1841 11197PERL_CONTEXT *
a8fc9800 11198Perl_cx_dup(pTHX_ PERL_CONTEXT *cxs, I32 ix, I32 max, CLONE_PARAMS* param)
1d7c1841
GS
11199{
11200 PERL_CONTEXT *ncxs;
11201
7918f24d
NC
11202 PERL_ARGS_ASSERT_CX_DUP;
11203
1d7c1841
GS
11204 if (!cxs)
11205 return (PERL_CONTEXT*)NULL;
11206
11207 /* look for it in the table first */
11208 ncxs = (PERL_CONTEXT*)ptr_table_fetch(PL_ptr_table, cxs);
11209 if (ncxs)
11210 return ncxs;
11211
11212 /* create anew and remember what it is */
c2d565bf 11213 Newx(ncxs, max + 1, PERL_CONTEXT);
1d7c1841 11214 ptr_table_store(PL_ptr_table, cxs, ncxs);
c2d565bf 11215 Copy(cxs, ncxs, max + 1, PERL_CONTEXT);
1d7c1841
GS
11216
11217 while (ix >= 0) {
c445ea15 11218 PERL_CONTEXT * const ncx = &ncxs[ix];
c2d565bf 11219 if (CxTYPE(ncx) == CXt_SUBST) {
1d7c1841
GS
11220 Perl_croak(aTHX_ "Cloning substitution context is unimplemented");
11221 }
11222 else {
c2d565bf 11223 switch (CxTYPE(ncx)) {
1d7c1841 11224 case CXt_SUB:
c2d565bf
NC
11225 ncx->blk_sub.cv = (ncx->blk_sub.olddepth == 0
11226 ? cv_dup_inc(ncx->blk_sub.cv, param)
11227 : cv_dup(ncx->blk_sub.cv,param));
bafb2adc 11228 ncx->blk_sub.argarray = (CxHASARGS(ncx)
c2d565bf
NC
11229 ? av_dup_inc(ncx->blk_sub.argarray,
11230 param)
7d49f689 11231 : NULL);
c2d565bf
NC
11232 ncx->blk_sub.savearray = av_dup_inc(ncx->blk_sub.savearray,
11233 param);
d8d97e70 11234 ncx->blk_sub.oldcomppad = (PAD*)ptr_table_fetch(PL_ptr_table,
c2d565bf 11235 ncx->blk_sub.oldcomppad);
1d7c1841
GS
11236 break;
11237 case CXt_EVAL:
c2d565bf
NC
11238 ncx->blk_eval.old_namesv = sv_dup_inc(ncx->blk_eval.old_namesv,
11239 param);
11240 ncx->blk_eval.cur_text = sv_dup(ncx->blk_eval.cur_text, param);
1d7c1841 11241 break;
d01136d6 11242 case CXt_LOOP_LAZYSV:
d01136d6
BS
11243 ncx->blk_loop.state_u.lazysv.end
11244 = sv_dup_inc(ncx->blk_loop.state_u.lazysv.end, param);
840fe433
NC
11245 /* We are taking advantage of av_dup_inc and sv_dup_inc
11246 actually being the same function, and order equivalance of
11247 the two unions.
11248 We can assert the later [but only at run time :-(] */
11249 assert ((void *) &ncx->blk_loop.state_u.ary.ary ==
11250 (void *) &ncx->blk_loop.state_u.lazysv.cur);
3b719c58 11251 case CXt_LOOP_FOR:
d01136d6
BS
11252 ncx->blk_loop.state_u.ary.ary
11253 = av_dup_inc(ncx->blk_loop.state_u.ary.ary, param);
11254 case CXt_LOOP_LAZYIV:
3b719c58 11255 case CXt_LOOP_PLAIN:
e846cb92
NC
11256 if (CxPADLOOP(ncx)) {
11257 ncx->blk_loop.oldcomppad
11258 = (PAD*)ptr_table_fetch(PL_ptr_table,
11259 ncx->blk_loop.oldcomppad);
11260 } else {
11261 ncx->blk_loop.oldcomppad
159b6efe
NC
11262 = (PAD*)gv_dup((const GV *)ncx->blk_loop.oldcomppad,
11263 param);
e846cb92 11264 }
1d7c1841
GS
11265 break;
11266 case CXt_FORMAT:
f9c764c5
NC
11267 ncx->blk_format.cv = cv_dup(ncx->blk_format.cv, param);
11268 ncx->blk_format.gv = gv_dup(ncx->blk_format.gv, param);
11269 ncx->blk_format.dfoutgv = gv_dup_inc(ncx->blk_format.dfoutgv,
c2d565bf 11270 param);
1d7c1841
GS
11271 break;
11272 case CXt_BLOCK:
11273 case CXt_NULL:
11274 break;
11275 }
11276 }
11277 --ix;
11278 }
11279 return ncxs;
11280}
11281
645c22ef
DM
11282/* duplicate a stack info structure */
11283
1d7c1841 11284PERL_SI *
a8fc9800 11285Perl_si_dup(pTHX_ PERL_SI *si, CLONE_PARAMS* param)
1d7c1841
GS
11286{
11287 PERL_SI *nsi;
11288
7918f24d
NC
11289 PERL_ARGS_ASSERT_SI_DUP;
11290
1d7c1841
GS
11291 if (!si)
11292 return (PERL_SI*)NULL;
11293
11294 /* look for it in the table first */
11295 nsi = (PERL_SI*)ptr_table_fetch(PL_ptr_table, si);
11296 if (nsi)
11297 return nsi;
11298
11299 /* create anew and remember what it is */
a02a5408 11300 Newxz(nsi, 1, PERL_SI);
1d7c1841
GS
11301 ptr_table_store(PL_ptr_table, si, nsi);
11302
d2d73c3e 11303 nsi->si_stack = av_dup_inc(si->si_stack, param);
1d7c1841
GS
11304 nsi->si_cxix = si->si_cxix;
11305 nsi->si_cxmax = si->si_cxmax;
d2d73c3e 11306 nsi->si_cxstack = cx_dup(si->si_cxstack, si->si_cxix, si->si_cxmax, param);
1d7c1841 11307 nsi->si_type = si->si_type;
d2d73c3e
AB
11308 nsi->si_prev = si_dup(si->si_prev, param);
11309 nsi->si_next = si_dup(si->si_next, param);
1d7c1841
GS
11310 nsi->si_markoff = si->si_markoff;
11311
11312 return nsi;
11313}
11314
11315#define POPINT(ss,ix) ((ss)[--(ix)].any_i32)
11316#define TOPINT(ss,ix) ((ss)[ix].any_i32)
11317#define POPLONG(ss,ix) ((ss)[--(ix)].any_long)
11318#define TOPLONG(ss,ix) ((ss)[ix].any_long)
11319#define POPIV(ss,ix) ((ss)[--(ix)].any_iv)
11320#define TOPIV(ss,ix) ((ss)[ix].any_iv)
38d8b13e
HS
11321#define POPBOOL(ss,ix) ((ss)[--(ix)].any_bool)
11322#define TOPBOOL(ss,ix) ((ss)[ix].any_bool)
1d7c1841
GS
11323#define POPPTR(ss,ix) ((ss)[--(ix)].any_ptr)
11324#define TOPPTR(ss,ix) ((ss)[ix].any_ptr)
11325#define POPDPTR(ss,ix) ((ss)[--(ix)].any_dptr)
11326#define TOPDPTR(ss,ix) ((ss)[ix].any_dptr)
11327#define POPDXPTR(ss,ix) ((ss)[--(ix)].any_dxptr)
11328#define TOPDXPTR(ss,ix) ((ss)[ix].any_dxptr)
11329
11330/* XXXXX todo */
11331#define pv_dup_inc(p) SAVEPV(p)
11332#define pv_dup(p) SAVEPV(p)
11333#define svp_dup_inc(p,pp) any_dup(p,pp)
11334
645c22ef
DM
11335/* map any object to the new equivent - either something in the
11336 * ptr table, or something in the interpreter structure
11337 */
11338
1d7c1841 11339void *
53c1dcc0 11340Perl_any_dup(pTHX_ void *v, const PerlInterpreter *proto_perl)
1d7c1841
GS
11341{
11342 void *ret;
11343
7918f24d
NC
11344 PERL_ARGS_ASSERT_ANY_DUP;
11345
1d7c1841
GS
11346 if (!v)
11347 return (void*)NULL;
11348
11349 /* look for it in the table first */
11350 ret = ptr_table_fetch(PL_ptr_table, v);
11351 if (ret)
11352 return ret;
11353
11354 /* see if it is part of the interpreter structure */
11355 if (v >= (void*)proto_perl && v < (void*)(proto_perl+1))
acfe0abc 11356 ret = (void*)(((char*)aTHX) + (((char*)v) - (char*)proto_perl));
05ec9bb3 11357 else {
1d7c1841 11358 ret = v;
05ec9bb3 11359 }
1d7c1841
GS
11360
11361 return ret;
11362}
11363
645c22ef
DM
11364/* duplicate the save stack */
11365
1d7c1841 11366ANY *
a8fc9800 11367Perl_ss_dup(pTHX_ PerlInterpreter *proto_perl, CLONE_PARAMS* param)
1d7c1841 11368{
53d44271 11369 dVAR;
907b3e23
DM
11370 ANY * const ss = proto_perl->Isavestack;
11371 const I32 max = proto_perl->Isavestack_max;
11372 I32 ix = proto_perl->Isavestack_ix;
1d7c1841 11373 ANY *nss;
daba3364 11374 const SV *sv;
1d193675
NC
11375 const GV *gv;
11376 const AV *av;
11377 const HV *hv;
1d7c1841
GS
11378 void* ptr;
11379 int intval;
11380 long longval;
11381 GP *gp;
11382 IV iv;
b24356f5 11383 I32 i;
c4e33207 11384 char *c = NULL;
1d7c1841 11385 void (*dptr) (void*);
acfe0abc 11386 void (*dxptr) (pTHX_ void*);
1d7c1841 11387
7918f24d
NC
11388 PERL_ARGS_ASSERT_SS_DUP;
11389
a02a5408 11390 Newxz(nss, max, ANY);
1d7c1841
GS
11391
11392 while (ix > 0) {
b24356f5
NC
11393 const I32 type = POPINT(ss,ix);
11394 TOPINT(nss,ix) = type;
11395 switch (type) {
3e07292d 11396 case SAVEt_HELEM: /* hash element */
daba3364 11397 sv = (const SV *)POPPTR(ss,ix);
3e07292d
NC
11398 TOPPTR(nss,ix) = sv_dup_inc(sv, param);
11399 /* fall through */
1d7c1841 11400 case SAVEt_ITEM: /* normal string */
a41cc44e 11401 case SAVEt_SV: /* scalar reference */
daba3364 11402 sv = (const SV *)POPPTR(ss,ix);
d2d73c3e 11403 TOPPTR(nss,ix) = sv_dup_inc(sv, param);
3e07292d
NC
11404 /* fall through */
11405 case SAVEt_FREESV:
11406 case SAVEt_MORTALIZESV:
daba3364 11407 sv = (const SV *)POPPTR(ss,ix);
d2d73c3e 11408 TOPPTR(nss,ix) = sv_dup_inc(sv, param);
1d7c1841 11409 break;
05ec9bb3
NIS
11410 case SAVEt_SHARED_PVREF: /* char* in shared space */
11411 c = (char*)POPPTR(ss,ix);
11412 TOPPTR(nss,ix) = savesharedpv(c);
11413 ptr = POPPTR(ss,ix);
11414 TOPPTR(nss,ix) = any_dup(ptr, proto_perl);
11415 break;
1d7c1841
GS
11416 case SAVEt_GENERIC_SVREF: /* generic sv */
11417 case SAVEt_SVREF: /* scalar reference */
daba3364 11418 sv = (const SV *)POPPTR(ss,ix);
d2d73c3e 11419 TOPPTR(nss,ix) = sv_dup_inc(sv, param);
1d7c1841
GS
11420 ptr = POPPTR(ss,ix);
11421 TOPPTR(nss,ix) = svp_dup_inc((SV**)ptr, proto_perl);/* XXXXX */
11422 break;
a41cc44e 11423 case SAVEt_HV: /* hash reference */
1d7c1841 11424 case SAVEt_AV: /* array reference */
daba3364 11425 sv = (const SV *) POPPTR(ss,ix);
337d28f5 11426 TOPPTR(nss,ix) = sv_dup_inc(sv, param);
3e07292d
NC
11427 /* fall through */
11428 case SAVEt_COMPPAD:
11429 case SAVEt_NSTAB:
daba3364 11430 sv = (const SV *) POPPTR(ss,ix);
3e07292d 11431 TOPPTR(nss,ix) = sv_dup(sv, param);
1d7c1841
GS
11432 break;
11433 case SAVEt_INT: /* int reference */
11434 ptr = POPPTR(ss,ix);
11435 TOPPTR(nss,ix) = any_dup(ptr, proto_perl);
11436 intval = (int)POPINT(ss,ix);
11437 TOPINT(nss,ix) = intval;
11438 break;
11439 case SAVEt_LONG: /* long reference */
11440 ptr = POPPTR(ss,ix);
11441 TOPPTR(nss,ix) = any_dup(ptr, proto_perl);
3e07292d
NC
11442 /* fall through */
11443 case SAVEt_CLEARSV:
1d7c1841
GS
11444 longval = (long)POPLONG(ss,ix);
11445 TOPLONG(nss,ix) = longval;
11446 break;
11447 case SAVEt_I32: /* I32 reference */
11448 case SAVEt_I16: /* I16 reference */
11449 case SAVEt_I8: /* I8 reference */
88effcc9 11450 case SAVEt_COP_ARYBASE: /* call CopARYBASE_set */
1d7c1841
GS
11451 ptr = POPPTR(ss,ix);
11452 TOPPTR(nss,ix) = any_dup(ptr, proto_perl);
1ccabee8 11453 i = POPINT(ss,ix);
1d7c1841
GS
11454 TOPINT(nss,ix) = i;
11455 break;
11456 case SAVEt_IV: /* IV reference */
11457 ptr = POPPTR(ss,ix);
11458 TOPPTR(nss,ix) = any_dup(ptr, proto_perl);
11459 iv = POPIV(ss,ix);
11460 TOPIV(nss,ix) = iv;
11461 break;
a41cc44e
NC
11462 case SAVEt_HPTR: /* HV* reference */
11463 case SAVEt_APTR: /* AV* reference */
1d7c1841
GS
11464 case SAVEt_SPTR: /* SV* reference */
11465 ptr = POPPTR(ss,ix);
11466 TOPPTR(nss,ix) = any_dup(ptr, proto_perl);
daba3364 11467 sv = (const SV *)POPPTR(ss,ix);
d2d73c3e 11468 TOPPTR(nss,ix) = sv_dup(sv, param);
1d7c1841
GS
11469 break;
11470 case SAVEt_VPTR: /* random* reference */
11471 ptr = POPPTR(ss,ix);
11472 TOPPTR(nss,ix) = any_dup(ptr, proto_perl);
11473 ptr = POPPTR(ss,ix);
11474 TOPPTR(nss,ix) = any_dup(ptr, proto_perl);
11475 break;
b03d03b0 11476 case SAVEt_GENERIC_PVREF: /* generic char* */
1d7c1841
GS
11477 case SAVEt_PPTR: /* char* reference */
11478 ptr = POPPTR(ss,ix);
11479 TOPPTR(nss,ix) = any_dup(ptr, proto_perl);
11480 c = (char*)POPPTR(ss,ix);
11481 TOPPTR(nss,ix) = pv_dup(c);
11482 break;
1d7c1841
GS
11483 case SAVEt_GP: /* scalar reference */
11484 gp = (GP*)POPPTR(ss,ix);
d2d73c3e 11485 TOPPTR(nss,ix) = gp = gp_dup(gp, param);
1d7c1841 11486 (void)GpREFCNT_inc(gp);
159b6efe 11487 gv = (const GV *)POPPTR(ss,ix);
2ed3c8fc 11488 TOPPTR(nss,ix) = gv_dup_inc(gv, param);
1d7c1841 11489 break;
1d7c1841
GS
11490 case SAVEt_FREEOP:
11491 ptr = POPPTR(ss,ix);
11492 if (ptr && (((OP*)ptr)->op_private & OPpREFCOUNTED)) {
11493 /* these are assumed to be refcounted properly */
53c1dcc0 11494 OP *o;
1d7c1841
GS
11495 switch (((OP*)ptr)->op_type) {
11496 case OP_LEAVESUB:
11497 case OP_LEAVESUBLV:
11498 case OP_LEAVEEVAL:
11499 case OP_LEAVE:
11500 case OP_SCOPE:
11501 case OP_LEAVEWRITE:
e977893f
GS
11502 TOPPTR(nss,ix) = ptr;
11503 o = (OP*)ptr;
d3c72c2a 11504 OP_REFCNT_LOCK;
594cd643 11505 (void) OpREFCNT_inc(o);
d3c72c2a 11506 OP_REFCNT_UNLOCK;
1d7c1841
GS
11507 break;
11508 default:
5f66b61c 11509 TOPPTR(nss,ix) = NULL;
1d7c1841
GS
11510 break;
11511 }
11512 }
11513 else
5f66b61c 11514 TOPPTR(nss,ix) = NULL;
1d7c1841 11515 break;
1d7c1841 11516 case SAVEt_DELETE:
1d193675 11517 hv = (const HV *)POPPTR(ss,ix);
d2d73c3e 11518 TOPPTR(nss,ix) = hv_dup_inc(hv, param);
35d4f826
NC
11519 i = POPINT(ss,ix);
11520 TOPINT(nss,ix) = i;
8e41545f
NC
11521 /* Fall through */
11522 case SAVEt_FREEPV:
1d7c1841
GS
11523 c = (char*)POPPTR(ss,ix);
11524 TOPPTR(nss,ix) = pv_dup_inc(c);
35d4f826 11525 break;
3e07292d 11526 case SAVEt_STACK_POS: /* Position on Perl stack */
1d7c1841
GS
11527 i = POPINT(ss,ix);
11528 TOPINT(nss,ix) = i;
11529 break;
11530 case SAVEt_DESTRUCTOR:
11531 ptr = POPPTR(ss,ix);
11532 TOPPTR(nss,ix) = any_dup(ptr, proto_perl); /* XXX quite arbitrary */
11533 dptr = POPDPTR(ss,ix);
8141890a
JH
11534 TOPDPTR(nss,ix) = DPTR2FPTR(void (*)(void*),
11535 any_dup(FPTR2DPTR(void *, dptr),
11536 proto_perl));
1d7c1841
GS
11537 break;
11538 case SAVEt_DESTRUCTOR_X:
11539 ptr = POPPTR(ss,ix);
11540 TOPPTR(nss,ix) = any_dup(ptr, proto_perl); /* XXX quite arbitrary */
11541 dxptr = POPDXPTR(ss,ix);
8141890a
JH
11542 TOPDXPTR(nss,ix) = DPTR2FPTR(void (*)(pTHX_ void*),
11543 any_dup(FPTR2DPTR(void *, dxptr),
11544 proto_perl));
1d7c1841
GS
11545 break;
11546 case SAVEt_REGCONTEXT:
11547 case SAVEt_ALLOC:
11548 i = POPINT(ss,ix);
11549 TOPINT(nss,ix) = i;
11550 ix -= i;
11551 break;
1d7c1841 11552 case SAVEt_AELEM: /* array element */
daba3364 11553 sv = (const SV *)POPPTR(ss,ix);
d2d73c3e 11554 TOPPTR(nss,ix) = sv_dup_inc(sv, param);
1d7c1841
GS
11555 i = POPINT(ss,ix);
11556 TOPINT(nss,ix) = i;
502c6561 11557 av = (const AV *)POPPTR(ss,ix);
d2d73c3e 11558 TOPPTR(nss,ix) = av_dup_inc(av, param);
1d7c1841 11559 break;
1d7c1841
GS
11560 case SAVEt_OP:
11561 ptr = POPPTR(ss,ix);
11562 TOPPTR(nss,ix) = ptr;
11563 break;
11564 case SAVEt_HINTS:
b3ca2e83 11565 ptr = POPPTR(ss,ix);
080ac856 11566 if (ptr) {
7b6dd8c3 11567 HINTS_REFCNT_LOCK;
080ac856 11568 ((struct refcounted_he *)ptr)->refcounted_he_refcnt++;
7b6dd8c3
NC
11569 HINTS_REFCNT_UNLOCK;
11570 }
cbb1fbea 11571 TOPPTR(nss,ix) = ptr;
601cee3b
NC
11572 i = POPINT(ss,ix);
11573 TOPINT(nss,ix) = i;
a8f8b6a7 11574 if (i & HINT_LOCALIZE_HH) {
1d193675 11575 hv = (const HV *)POPPTR(ss,ix);
a8f8b6a7
NC
11576 TOPPTR(nss,ix) = hv_dup_inc(hv, param);
11577 }
1d7c1841 11578 break;
09edbca0 11579 case SAVEt_PADSV_AND_MORTALIZE:
c3564e5c
GS
11580 longval = (long)POPLONG(ss,ix);
11581 TOPLONG(nss,ix) = longval;
11582 ptr = POPPTR(ss,ix);
11583 TOPPTR(nss,ix) = any_dup(ptr, proto_perl);
daba3364 11584 sv = (const SV *)POPPTR(ss,ix);
09edbca0 11585 TOPPTR(nss,ix) = sv_dup_inc(sv, param);
c3564e5c 11586 break;
a1bb4754 11587 case SAVEt_BOOL:
38d8b13e 11588 ptr = POPPTR(ss,ix);
b9609c01 11589 TOPPTR(nss,ix) = any_dup(ptr, proto_perl);
38d8b13e 11590 longval = (long)POPBOOL(ss,ix);
b9609c01 11591 TOPBOOL(nss,ix) = (bool)longval;
a1bb4754 11592 break;
8bd2680e
MHM
11593 case SAVEt_SET_SVFLAGS:
11594 i = POPINT(ss,ix);
11595 TOPINT(nss,ix) = i;
11596 i = POPINT(ss,ix);
11597 TOPINT(nss,ix) = i;
daba3364 11598 sv = (const SV *)POPPTR(ss,ix);
8bd2680e
MHM
11599 TOPPTR(nss,ix) = sv_dup(sv, param);
11600 break;
5bfb7d0e
NC
11601 case SAVEt_RE_STATE:
11602 {
11603 const struct re_save_state *const old_state
11604 = (struct re_save_state *)
11605 (ss + ix - SAVESTACK_ALLOC_FOR_RE_SAVE_STATE);
11606 struct re_save_state *const new_state
11607 = (struct re_save_state *)
11608 (nss + ix - SAVESTACK_ALLOC_FOR_RE_SAVE_STATE);
11609
11610 Copy(old_state, new_state, 1, struct re_save_state);
11611 ix -= SAVESTACK_ALLOC_FOR_RE_SAVE_STATE;
11612
11613 new_state->re_state_bostr
11614 = pv_dup(old_state->re_state_bostr);
11615 new_state->re_state_reginput
11616 = pv_dup(old_state->re_state_reginput);
5bfb7d0e
NC
11617 new_state->re_state_regeol
11618 = pv_dup(old_state->re_state_regeol);
f0ab9afb
NC
11619 new_state->re_state_regoffs
11620 = (regexp_paren_pair*)
11621 any_dup(old_state->re_state_regoffs, proto_perl);
5bfb7d0e 11622 new_state->re_state_reglastparen
11b79775
DD
11623 = (U32*) any_dup(old_state->re_state_reglastparen,
11624 proto_perl);
5bfb7d0e 11625 new_state->re_state_reglastcloseparen
11b79775 11626 = (U32*)any_dup(old_state->re_state_reglastcloseparen,
5bfb7d0e 11627 proto_perl);
5bfb7d0e
NC
11628 /* XXX This just has to be broken. The old save_re_context
11629 code did SAVEGENERICPV(PL_reg_start_tmp);
11630 PL_reg_start_tmp is char **.
11631 Look above to what the dup code does for
11632 SAVEt_GENERIC_PVREF
11633 It can never have worked.
11634 So this is merely a faithful copy of the exiting bug: */
11635 new_state->re_state_reg_start_tmp
11636 = (char **) pv_dup((char *)
11637 old_state->re_state_reg_start_tmp);
11638 /* I assume that it only ever "worked" because no-one called
11639 (pseudo)fork while the regexp engine had re-entered itself.
11640 */
5bfb7d0e
NC
11641#ifdef PERL_OLD_COPY_ON_WRITE
11642 new_state->re_state_nrs
11643 = sv_dup(old_state->re_state_nrs, param);
11644#endif
11645 new_state->re_state_reg_magic
11b79775
DD
11646 = (MAGIC*) any_dup(old_state->re_state_reg_magic,
11647 proto_perl);
5bfb7d0e 11648 new_state->re_state_reg_oldcurpm
11b79775
DD
11649 = (PMOP*) any_dup(old_state->re_state_reg_oldcurpm,
11650 proto_perl);
5bfb7d0e 11651 new_state->re_state_reg_curpm
11b79775
DD
11652 = (PMOP*) any_dup(old_state->re_state_reg_curpm,
11653 proto_perl);
5bfb7d0e
NC
11654 new_state->re_state_reg_oldsaved
11655 = pv_dup(old_state->re_state_reg_oldsaved);
11656 new_state->re_state_reg_poscache
11657 = pv_dup(old_state->re_state_reg_poscache);
5bfb7d0e
NC
11658 new_state->re_state_reg_starttry
11659 = pv_dup(old_state->re_state_reg_starttry);
5bfb7d0e
NC
11660 break;
11661 }
68da3b2f
NC
11662 case SAVEt_COMPILE_WARNINGS:
11663 ptr = POPPTR(ss,ix);
11664 TOPPTR(nss,ix) = DUP_WARNINGS((STRLEN*)ptr);
7b6dd8c3 11665 break;
7c197c94
DM
11666 case SAVEt_PARSER:
11667 ptr = POPPTR(ss,ix);
456084a8 11668 TOPPTR(nss,ix) = parser_dup((const yy_parser*)ptr, param);
7c197c94 11669 break;
1d7c1841 11670 default:
147bc374
NC
11671 Perl_croak(aTHX_
11672 "panic: ss_dup inconsistency (%"IVdf")", (IV) type);
1d7c1841
GS
11673 }
11674 }
11675
bd81e77b
NC
11676 return nss;
11677}
11678
11679
11680/* if sv is a stash, call $class->CLONE_SKIP(), and set the SVphv_CLONEABLE
11681 * flag to the result. This is done for each stash before cloning starts,
11682 * so we know which stashes want their objects cloned */
11683
11684static void
f30de749 11685do_mark_cloneable_stash(pTHX_ SV *const sv)
bd81e77b 11686{
1d193675 11687 const HEK * const hvname = HvNAME_HEK((const HV *)sv);
bd81e77b 11688 if (hvname) {
85fbaab2 11689 GV* const cloner = gv_fetchmethod_autoload(MUTABLE_HV(sv), "CLONE_SKIP", 0);
bd81e77b
NC
11690 SvFLAGS(sv) |= SVphv_CLONEABLE; /* clone objects by default */
11691 if (cloner && GvCV(cloner)) {
11692 dSP;
11693 UV status;
11694
11695 ENTER;
11696 SAVETMPS;
11697 PUSHMARK(SP);
6e449a3a 11698 mXPUSHs(newSVhek(hvname));
bd81e77b 11699 PUTBACK;
daba3364 11700 call_sv(MUTABLE_SV(GvCV(cloner)), G_SCALAR);
bd81e77b
NC
11701 SPAGAIN;
11702 status = POPu;
11703 PUTBACK;
11704 FREETMPS;
11705 LEAVE;
11706 if (status)
11707 SvFLAGS(sv) &= ~SVphv_CLONEABLE;
11708 }
11709 }
11710}
11711
11712
11713
11714/*
11715=for apidoc perl_clone
11716
11717Create and return a new interpreter by cloning the current one.
11718
11719perl_clone takes these flags as parameters:
11720
11721CLONEf_COPY_STACKS - is used to, well, copy the stacks also,
11722without it we only clone the data and zero the stacks,
11723with it we copy the stacks and the new perl interpreter is
11724ready to run at the exact same point as the previous one.
11725The pseudo-fork code uses COPY_STACKS while the
878090d5 11726threads->create doesn't.
bd81e77b
NC
11727
11728CLONEf_KEEP_PTR_TABLE
11729perl_clone keeps a ptr_table with the pointer of the old
11730variable as a key and the new variable as a value,
11731this allows it to check if something has been cloned and not
11732clone it again but rather just use the value and increase the
11733refcount. If KEEP_PTR_TABLE is not set then perl_clone will kill
11734the ptr_table using the function
11735C<ptr_table_free(PL_ptr_table); PL_ptr_table = NULL;>,
11736reason to keep it around is if you want to dup some of your own
11737variable who are outside the graph perl scans, example of this
11738code is in threads.xs create
11739
11740CLONEf_CLONE_HOST
11741This is a win32 thing, it is ignored on unix, it tells perls
11742win32host code (which is c++) to clone itself, this is needed on
11743win32 if you want to run two threads at the same time,
11744if you just want to do some stuff in a separate perl interpreter
11745and then throw it away and return to the original one,
11746you don't need to do anything.
11747
11748=cut
11749*/
11750
11751/* XXX the above needs expanding by someone who actually understands it ! */
11752EXTERN_C PerlInterpreter *
11753perl_clone_host(PerlInterpreter* proto_perl, UV flags);
11754
11755PerlInterpreter *
11756perl_clone(PerlInterpreter *proto_perl, UV flags)
11757{
11758 dVAR;
11759#ifdef PERL_IMPLICIT_SYS
11760
7918f24d
NC
11761 PERL_ARGS_ASSERT_PERL_CLONE;
11762
bd81e77b
NC
11763 /* perlhost.h so we need to call into it
11764 to clone the host, CPerlHost should have a c interface, sky */
11765
11766 if (flags & CLONEf_CLONE_HOST) {
11767 return perl_clone_host(proto_perl,flags);
11768 }
11769 return perl_clone_using(proto_perl, flags,
11770 proto_perl->IMem,
11771 proto_perl->IMemShared,
11772 proto_perl->IMemParse,
11773 proto_perl->IEnv,
11774 proto_perl->IStdIO,
11775 proto_perl->ILIO,
11776 proto_perl->IDir,
11777 proto_perl->ISock,
11778 proto_perl->IProc);
11779}
11780
11781PerlInterpreter *
11782perl_clone_using(PerlInterpreter *proto_perl, UV flags,
11783 struct IPerlMem* ipM, struct IPerlMem* ipMS,
11784 struct IPerlMem* ipMP, struct IPerlEnv* ipE,
11785 struct IPerlStdIO* ipStd, struct IPerlLIO* ipLIO,
11786 struct IPerlDir* ipD, struct IPerlSock* ipS,
11787 struct IPerlProc* ipP)
11788{
11789 /* XXX many of the string copies here can be optimized if they're
11790 * constants; they need to be allocated as common memory and just
11791 * their pointers copied. */
11792
11793 IV i;
11794 CLONE_PARAMS clone_params;
5f66b61c 11795 CLONE_PARAMS* const param = &clone_params;
bd81e77b 11796
5f66b61c 11797 PerlInterpreter * const my_perl = (PerlInterpreter*)(*ipM->pMalloc)(ipM, sizeof(PerlInterpreter));
7918f24d
NC
11798
11799 PERL_ARGS_ASSERT_PERL_CLONE_USING;
11800
bd81e77b
NC
11801 /* for each stash, determine whether its objects should be cloned */
11802 S_visit(proto_perl, do_mark_cloneable_stash, SVt_PVHV, SVTYPEMASK);
11803 PERL_SET_THX(my_perl);
11804
11805# ifdef DEBUGGING
7e337ee0 11806 PoisonNew(my_perl, 1, PerlInterpreter);
5f66b61c
AL
11807 PL_op = NULL;
11808 PL_curcop = NULL;
bd81e77b
NC
11809 PL_markstack = 0;
11810 PL_scopestack = 0;
cbdd5331 11811 PL_scopestack_name = 0;
bd81e77b
NC
11812 PL_savestack = 0;
11813 PL_savestack_ix = 0;
11814 PL_savestack_max = -1;
11815 PL_sig_pending = 0;
b8328dae 11816 PL_parser = NULL;
bd81e77b
NC
11817 Zero(&PL_debug_pad, 1, struct perl_debug_pad);
11818# else /* !DEBUGGING */
11819 Zero(my_perl, 1, PerlInterpreter);
11820# endif /* DEBUGGING */
11821
11822 /* host pointers */
11823 PL_Mem = ipM;
11824 PL_MemShared = ipMS;
11825 PL_MemParse = ipMP;
11826 PL_Env = ipE;
11827 PL_StdIO = ipStd;
11828 PL_LIO = ipLIO;
11829 PL_Dir = ipD;
11830 PL_Sock = ipS;
11831 PL_Proc = ipP;
11832#else /* !PERL_IMPLICIT_SYS */
11833 IV i;
11834 CLONE_PARAMS clone_params;
11835 CLONE_PARAMS* param = &clone_params;
5f66b61c 11836 PerlInterpreter * const my_perl = (PerlInterpreter*)PerlMem_malloc(sizeof(PerlInterpreter));
7918f24d
NC
11837
11838 PERL_ARGS_ASSERT_PERL_CLONE;
11839
bd81e77b
NC
11840 /* for each stash, determine whether its objects should be cloned */
11841 S_visit(proto_perl, do_mark_cloneable_stash, SVt_PVHV, SVTYPEMASK);
11842 PERL_SET_THX(my_perl);
11843
11844# ifdef DEBUGGING
7e337ee0 11845 PoisonNew(my_perl, 1, PerlInterpreter);
5f66b61c
AL
11846 PL_op = NULL;
11847 PL_curcop = NULL;
bd81e77b
NC
11848 PL_markstack = 0;
11849 PL_scopestack = 0;
cbdd5331 11850 PL_scopestack_name = 0;
bd81e77b
NC
11851 PL_savestack = 0;
11852 PL_savestack_ix = 0;
11853 PL_savestack_max = -1;
11854 PL_sig_pending = 0;
b8328dae 11855 PL_parser = NULL;
bd81e77b
NC
11856 Zero(&PL_debug_pad, 1, struct perl_debug_pad);
11857# else /* !DEBUGGING */
11858 Zero(my_perl, 1, PerlInterpreter);
11859# endif /* DEBUGGING */
11860#endif /* PERL_IMPLICIT_SYS */
11861 param->flags = flags;
11862 param->proto_perl = proto_perl;
11863
7cb608b5
NC
11864 INIT_TRACK_MEMPOOL(my_perl->Imemory_debug_header, my_perl);
11865
fdda85ca 11866 PL_body_arenas = NULL;
bd81e77b
NC
11867 Zero(&PL_body_roots, 1, PL_body_roots);
11868
11869 PL_nice_chunk = NULL;
11870 PL_nice_chunk_size = 0;
11871 PL_sv_count = 0;
11872 PL_sv_objcount = 0;
a0714e2c
SS
11873 PL_sv_root = NULL;
11874 PL_sv_arenaroot = NULL;
bd81e77b
NC
11875
11876 PL_debug = proto_perl->Idebug;
11877
11878 PL_hash_seed = proto_perl->Ihash_seed;
11879 PL_rehash_seed = proto_perl->Irehash_seed;
11880
11881#ifdef USE_REENTRANT_API
11882 /* XXX: things like -Dm will segfault here in perlio, but doing
11883 * PERL_SET_CONTEXT(proto_perl);
11884 * breaks too many other things
11885 */
11886 Perl_reentrant_init(aTHX);
11887#endif
11888
11889 /* create SV map for pointer relocation */
11890 PL_ptr_table = ptr_table_new();
11891
11892 /* initialize these special pointers as early as possible */
11893 SvANY(&PL_sv_undef) = NULL;
11894 SvREFCNT(&PL_sv_undef) = (~(U32)0)/2;
11895 SvFLAGS(&PL_sv_undef) = SVf_READONLY|SVt_NULL;
11896 ptr_table_store(PL_ptr_table, &proto_perl->Isv_undef, &PL_sv_undef);
11897
11898 SvANY(&PL_sv_no) = new_XPVNV();
11899 SvREFCNT(&PL_sv_no) = (~(U32)0)/2;
11900 SvFLAGS(&PL_sv_no) = SVp_IOK|SVf_IOK|SVp_NOK|SVf_NOK
11901 |SVp_POK|SVf_POK|SVf_READONLY|SVt_PVNV;
bb7a0f54 11902 SvPV_set(&PL_sv_no, savepvn(PL_No, 0));
bd81e77b
NC
11903 SvCUR_set(&PL_sv_no, 0);
11904 SvLEN_set(&PL_sv_no, 1);
11905 SvIV_set(&PL_sv_no, 0);
11906 SvNV_set(&PL_sv_no, 0);
11907 ptr_table_store(PL_ptr_table, &proto_perl->Isv_no, &PL_sv_no);
11908
11909 SvANY(&PL_sv_yes) = new_XPVNV();
11910 SvREFCNT(&PL_sv_yes) = (~(U32)0)/2;
11911 SvFLAGS(&PL_sv_yes) = SVp_IOK|SVf_IOK|SVp_NOK|SVf_NOK
11912 |SVp_POK|SVf_POK|SVf_READONLY|SVt_PVNV;
bb7a0f54 11913 SvPV_set(&PL_sv_yes, savepvn(PL_Yes, 1));
bd81e77b
NC
11914 SvCUR_set(&PL_sv_yes, 1);
11915 SvLEN_set(&PL_sv_yes, 2);
11916 SvIV_set(&PL_sv_yes, 1);
11917 SvNV_set(&PL_sv_yes, 1);
11918 ptr_table_store(PL_ptr_table, &proto_perl->Isv_yes, &PL_sv_yes);
11919
11920 /* create (a non-shared!) shared string table */
11921 PL_strtab = newHV();
11922 HvSHAREKEYS_off(PL_strtab);
11923 hv_ksplit(PL_strtab, HvTOTALKEYS(proto_perl->Istrtab));
11924 ptr_table_store(PL_ptr_table, proto_perl->Istrtab, PL_strtab);
11925
11926 PL_compiling = proto_perl->Icompiling;
11927
11928 /* These two PVs will be free'd special way so must set them same way op.c does */
11929 PL_compiling.cop_stashpv = savesharedpv(PL_compiling.cop_stashpv);
11930 ptr_table_store(PL_ptr_table, proto_perl->Icompiling.cop_stashpv, PL_compiling.cop_stashpv);
11931
11932 PL_compiling.cop_file = savesharedpv(PL_compiling.cop_file);
11933 ptr_table_store(PL_ptr_table, proto_perl->Icompiling.cop_file, PL_compiling.cop_file);
11934
11935 ptr_table_store(PL_ptr_table, &proto_perl->Icompiling, &PL_compiling);
72dc9ed5 11936 PL_compiling.cop_warnings = DUP_WARNINGS(PL_compiling.cop_warnings);
c28fe1ec 11937 if (PL_compiling.cop_hints_hash) {
cbb1fbea 11938 HINTS_REFCNT_LOCK;
c28fe1ec 11939 PL_compiling.cop_hints_hash->refcounted_he_refcnt++;
cbb1fbea
NC
11940 HINTS_REFCNT_UNLOCK;
11941 }
907b3e23 11942 PL_curcop = (COP*)any_dup(proto_perl->Icurcop, proto_perl);
5892a4d4
NC
11943#ifdef PERL_DEBUG_READONLY_OPS
11944 PL_slabs = NULL;
11945 PL_slab_count = 0;
11946#endif
bd81e77b
NC
11947
11948 /* pseudo environmental stuff */
11949 PL_origargc = proto_perl->Iorigargc;
11950 PL_origargv = proto_perl->Iorigargv;
11951
11952 param->stashes = newAV(); /* Setup array of objects to call clone on */
11953
11954 /* Set tainting stuff before PerlIO_debug can possibly get called */
11955 PL_tainting = proto_perl->Itainting;
11956 PL_taint_warn = proto_perl->Itaint_warn;
11957
11958#ifdef PERLIO_LAYERS
11959 /* Clone PerlIO tables as soon as we can handle general xx_dup() */
11960 PerlIO_clone(aTHX_ proto_perl, param);
11961#endif
11962
11963 PL_envgv = gv_dup(proto_perl->Ienvgv, param);
11964 PL_incgv = gv_dup(proto_perl->Iincgv, param);
11965 PL_hintgv = gv_dup(proto_perl->Ihintgv, param);
11966 PL_origfilename = SAVEPV(proto_perl->Iorigfilename);
11967 PL_diehook = sv_dup_inc(proto_perl->Idiehook, param);
11968 PL_warnhook = sv_dup_inc(proto_perl->Iwarnhook, param);
11969
11970 /* switches */
11971 PL_minus_c = proto_perl->Iminus_c;
11972 PL_patchlevel = sv_dup_inc(proto_perl->Ipatchlevel, param);
11973 PL_localpatches = proto_perl->Ilocalpatches;
11974 PL_splitstr = proto_perl->Isplitstr;
bd81e77b
NC
11975 PL_minus_n = proto_perl->Iminus_n;
11976 PL_minus_p = proto_perl->Iminus_p;
11977 PL_minus_l = proto_perl->Iminus_l;
11978 PL_minus_a = proto_perl->Iminus_a;
bc9b29db 11979 PL_minus_E = proto_perl->Iminus_E;
bd81e77b
NC
11980 PL_minus_F = proto_perl->Iminus_F;
11981 PL_doswitches = proto_perl->Idoswitches;
11982 PL_dowarn = proto_perl->Idowarn;
11983 PL_doextract = proto_perl->Idoextract;
11984 PL_sawampersand = proto_perl->Isawampersand;
11985 PL_unsafe = proto_perl->Iunsafe;
11986 PL_inplace = SAVEPV(proto_perl->Iinplace);
11987 PL_e_script = sv_dup_inc(proto_perl->Ie_script, param);
11988 PL_perldb = proto_perl->Iperldb;
11989 PL_perl_destruct_level = proto_perl->Iperl_destruct_level;
11990 PL_exit_flags = proto_perl->Iexit_flags;
11991
11992 /* magical thingies */
11993 /* XXX time(&PL_basetime) when asked for? */
11994 PL_basetime = proto_perl->Ibasetime;
11995 PL_formfeed = sv_dup(proto_perl->Iformfeed, param);
11996
11997 PL_maxsysfd = proto_perl->Imaxsysfd;
bd81e77b
NC
11998 PL_statusvalue = proto_perl->Istatusvalue;
11999#ifdef VMS
12000 PL_statusvalue_vms = proto_perl->Istatusvalue_vms;
12001#else
12002 PL_statusvalue_posix = proto_perl->Istatusvalue_posix;
12003#endif
12004 PL_encoding = sv_dup(proto_perl->Iencoding, param);
12005
76f68e9b
MHM
12006 sv_setpvs(PERL_DEBUG_PAD(0), ""); /* For regex debugging. */
12007 sv_setpvs(PERL_DEBUG_PAD(1), ""); /* ext/re needs these */
12008 sv_setpvs(PERL_DEBUG_PAD(2), ""); /* even without DEBUGGING. */
bd81e77b 12009
84da74a7 12010
f9f4320a 12011 /* RE engine related */
84da74a7
YO
12012 Zero(&PL_reg_state, 1, struct re_save_state);
12013 PL_reginterp_cnt = 0;
12014 PL_regmatch_slab = NULL;
12015
bd81e77b 12016 /* Clone the regex array */
937c6efd
NC
12017 /* ORANGE FIXME for plugins, probably in the SV dup code.
12018 newSViv(PTR2IV(CALLREGDUPE(
12019 INT2PTR(REGEXP *, SvIVX(regex)), param))))
12020 */
12021 PL_regex_padav = av_dup_inc(proto_perl->Iregex_padav, param);
bd81e77b
NC
12022 PL_regex_pad = AvARRAY(PL_regex_padav);
12023
12024 /* shortcuts to various I/O objects */
e23d9e2f 12025 PL_ofsgv = gv_dup(proto_perl->Iofsgv, param);
bd81e77b
NC
12026 PL_stdingv = gv_dup(proto_perl->Istdingv, param);
12027 PL_stderrgv = gv_dup(proto_perl->Istderrgv, param);
12028 PL_defgv = gv_dup(proto_perl->Idefgv, param);
12029 PL_argvgv = gv_dup(proto_perl->Iargvgv, param);
12030 PL_argvoutgv = gv_dup(proto_perl->Iargvoutgv, param);
12031 PL_argvout_stack = av_dup_inc(proto_perl->Iargvout_stack, param);
1d7c1841 12032
bd81e77b
NC
12033 /* shortcuts to regexp stuff */
12034 PL_replgv = gv_dup(proto_perl->Ireplgv, param);
9660f481 12035
bd81e77b
NC
12036 /* shortcuts to misc objects */
12037 PL_errgv = gv_dup(proto_perl->Ierrgv, param);
9660f481 12038
bd81e77b
NC
12039 /* shortcuts to debugging objects */
12040 PL_DBgv = gv_dup(proto_perl->IDBgv, param);
12041 PL_DBline = gv_dup(proto_perl->IDBline, param);
12042 PL_DBsub = gv_dup(proto_perl->IDBsub, param);
12043 PL_DBsingle = sv_dup(proto_perl->IDBsingle, param);
12044 PL_DBtrace = sv_dup(proto_perl->IDBtrace, param);
12045 PL_DBsignal = sv_dup(proto_perl->IDBsignal, param);
bd81e77b 12046 PL_dbargs = av_dup(proto_perl->Idbargs, param);
9660f481 12047
bd81e77b 12048 /* symbol tables */
907b3e23
DM
12049 PL_defstash = hv_dup_inc(proto_perl->Idefstash, param);
12050 PL_curstash = hv_dup(proto_perl->Icurstash, param);
bd81e77b
NC
12051 PL_debstash = hv_dup(proto_perl->Idebstash, param);
12052 PL_globalstash = hv_dup(proto_perl->Iglobalstash, param);
12053 PL_curstname = sv_dup_inc(proto_perl->Icurstname, param);
12054
12055 PL_beginav = av_dup_inc(proto_perl->Ibeginav, param);
12056 PL_beginav_save = av_dup_inc(proto_perl->Ibeginav_save, param);
12057 PL_checkav_save = av_dup_inc(proto_perl->Icheckav_save, param);
3c10abe3
AG
12058 PL_unitcheckav = av_dup_inc(proto_perl->Iunitcheckav, param);
12059 PL_unitcheckav_save = av_dup_inc(proto_perl->Iunitcheckav_save, param);
bd81e77b
NC
12060 PL_endav = av_dup_inc(proto_perl->Iendav, param);
12061 PL_checkav = av_dup_inc(proto_perl->Icheckav, param);
12062 PL_initav = av_dup_inc(proto_perl->Iinitav, param);
12063
12064 PL_sub_generation = proto_perl->Isub_generation;
dd69841b 12065 PL_isarev = hv_dup_inc(proto_perl->Iisarev, param);
bd81e77b
NC
12066
12067 /* funky return mechanisms */
12068 PL_forkprocess = proto_perl->Iforkprocess;
12069
12070 /* subprocess state */
12071 PL_fdpid = av_dup_inc(proto_perl->Ifdpid, param);
12072
12073 /* internal state */
12074 PL_maxo = proto_perl->Imaxo;
12075 if (proto_perl->Iop_mask)
12076 PL_op_mask = SAVEPVN(proto_perl->Iop_mask, PL_maxo);
12077 else
bd61b366 12078 PL_op_mask = NULL;
bd81e77b
NC
12079 /* PL_asserting = proto_perl->Iasserting; */
12080
12081 /* current interpreter roots */
12082 PL_main_cv = cv_dup_inc(proto_perl->Imain_cv, param);
d3c72c2a 12083 OP_REFCNT_LOCK;
bd81e77b 12084 PL_main_root = OpREFCNT_inc(proto_perl->Imain_root);
d3c72c2a 12085 OP_REFCNT_UNLOCK;
bd81e77b
NC
12086 PL_main_start = proto_perl->Imain_start;
12087 PL_eval_root = proto_perl->Ieval_root;
12088 PL_eval_start = proto_perl->Ieval_start;
12089
12090 /* runtime control stuff */
12091 PL_curcopdb = (COP*)any_dup(proto_perl->Icurcopdb, proto_perl);
bd81e77b
NC
12092
12093 PL_filemode = proto_perl->Ifilemode;
12094 PL_lastfd = proto_perl->Ilastfd;
12095 PL_oldname = proto_perl->Ioldname; /* XXX not quite right */
12096 PL_Argv = NULL;
bd61b366 12097 PL_Cmd = NULL;
bd81e77b 12098 PL_gensym = proto_perl->Igensym;
bd81e77b
NC
12099 PL_preambleav = av_dup_inc(proto_perl->Ipreambleav, param);
12100 PL_laststatval = proto_perl->Ilaststatval;
12101 PL_laststype = proto_perl->Ilaststype;
a0714e2c 12102 PL_mess_sv = NULL;
bd81e77b
NC
12103
12104 PL_ors_sv = sv_dup_inc(proto_perl->Iors_sv, param);
12105
12106 /* interpreter atexit processing */
12107 PL_exitlistlen = proto_perl->Iexitlistlen;
12108 if (PL_exitlistlen) {
12109 Newx(PL_exitlist, PL_exitlistlen, PerlExitListEntry);
12110 Copy(proto_perl->Iexitlist, PL_exitlist, PL_exitlistlen, PerlExitListEntry);
9660f481 12111 }
bd81e77b
NC
12112 else
12113 PL_exitlist = (PerlExitListEntry*)NULL;
f16dd614
DM
12114
12115 PL_my_cxt_size = proto_perl->Imy_cxt_size;
4c901e72 12116 if (PL_my_cxt_size) {
f16dd614
DM
12117 Newx(PL_my_cxt_list, PL_my_cxt_size, void *);
12118 Copy(proto_perl->Imy_cxt_list, PL_my_cxt_list, PL_my_cxt_size, void *);
53d44271 12119#ifdef PERL_GLOBAL_STRUCT_PRIVATE
bae1192d 12120 Newx(PL_my_cxt_keys, PL_my_cxt_size, const char *);
53d44271
JH
12121 Copy(proto_perl->Imy_cxt_keys, PL_my_cxt_keys, PL_my_cxt_size, char *);
12122#endif
f16dd614 12123 }
53d44271 12124 else {
f16dd614 12125 PL_my_cxt_list = (void**)NULL;
53d44271 12126#ifdef PERL_GLOBAL_STRUCT_PRIVATE
bae1192d 12127 PL_my_cxt_keys = (const char**)NULL;
53d44271
JH
12128#endif
12129 }
bd81e77b
NC
12130 PL_modglobal = hv_dup_inc(proto_perl->Imodglobal, param);
12131 PL_custom_op_names = hv_dup_inc(proto_perl->Icustom_op_names,param);
12132 PL_custom_op_descs = hv_dup_inc(proto_perl->Icustom_op_descs,param);
12133
12134 PL_profiledata = NULL;
9660f481 12135
bd81e77b 12136 PL_compcv = cv_dup(proto_perl->Icompcv, param);
9660f481 12137
bd81e77b 12138 PAD_CLONE_VARS(proto_perl, param);
9660f481 12139
bd81e77b
NC
12140#ifdef HAVE_INTERP_INTERN
12141 sys_intern_dup(&proto_perl->Isys_intern, &PL_sys_intern);
12142#endif
645c22ef 12143
bd81e77b
NC
12144 /* more statics moved here */
12145 PL_generation = proto_perl->Igeneration;
12146 PL_DBcv = cv_dup(proto_perl->IDBcv, param);
645c22ef 12147
bd81e77b
NC
12148 PL_in_clean_objs = proto_perl->Iin_clean_objs;
12149 PL_in_clean_all = proto_perl->Iin_clean_all;
6a78b4db 12150
bd81e77b
NC
12151 PL_uid = proto_perl->Iuid;
12152 PL_euid = proto_perl->Ieuid;
12153 PL_gid = proto_perl->Igid;
12154 PL_egid = proto_perl->Iegid;
12155 PL_nomemok = proto_perl->Inomemok;
12156 PL_an = proto_perl->Ian;
12157 PL_evalseq = proto_perl->Ievalseq;
12158 PL_origenviron = proto_perl->Iorigenviron; /* XXX not quite right */
12159 PL_origalen = proto_perl->Iorigalen;
12160#ifdef PERL_USES_PL_PIDSTATUS
12161 PL_pidstatus = newHV(); /* XXX flag for cloning? */
12162#endif
12163 PL_osname = SAVEPV(proto_perl->Iosname);
12164 PL_sighandlerp = proto_perl->Isighandlerp;
6a78b4db 12165
bd81e77b 12166 PL_runops = proto_perl->Irunops;
6a78b4db 12167
199e78b7
DM
12168 PL_parser = parser_dup(proto_perl->Iparser, param);
12169
f0c5aa00
DM
12170 /* XXX this only works if the saved cop has already been cloned */
12171 if (proto_perl->Iparser) {
12172 PL_parser->saved_curcop = (COP*)any_dup(
12173 proto_perl->Iparser->saved_curcop,
12174 proto_perl);
12175 }
12176
bd81e77b
NC
12177 PL_subline = proto_perl->Isubline;
12178 PL_subname = sv_dup_inc(proto_perl->Isubname, param);
c43294b8 12179
bd81e77b
NC
12180#ifdef FCRYPT
12181 PL_cryptseen = proto_perl->Icryptseen;
12182#endif
1d7c1841 12183
bd81e77b 12184 PL_hints = proto_perl->Ihints;
1d7c1841 12185
bd81e77b 12186 PL_amagic_generation = proto_perl->Iamagic_generation;
d2d73c3e 12187
bd81e77b
NC
12188#ifdef USE_LOCALE_COLLATE
12189 PL_collation_ix = proto_perl->Icollation_ix;
12190 PL_collation_name = SAVEPV(proto_perl->Icollation_name);
12191 PL_collation_standard = proto_perl->Icollation_standard;
12192 PL_collxfrm_base = proto_perl->Icollxfrm_base;
12193 PL_collxfrm_mult = proto_perl->Icollxfrm_mult;
12194#endif /* USE_LOCALE_COLLATE */
1d7c1841 12195
bd81e77b
NC
12196#ifdef USE_LOCALE_NUMERIC
12197 PL_numeric_name = SAVEPV(proto_perl->Inumeric_name);
12198 PL_numeric_standard = proto_perl->Inumeric_standard;
12199 PL_numeric_local = proto_perl->Inumeric_local;
12200 PL_numeric_radix_sv = sv_dup_inc(proto_perl->Inumeric_radix_sv, param);
12201#endif /* !USE_LOCALE_NUMERIC */
1d7c1841 12202
bd81e77b
NC
12203 /* utf8 character classes */
12204 PL_utf8_alnum = sv_dup_inc(proto_perl->Iutf8_alnum, param);
bd81e77b
NC
12205 PL_utf8_ascii = sv_dup_inc(proto_perl->Iutf8_ascii, param);
12206 PL_utf8_alpha = sv_dup_inc(proto_perl->Iutf8_alpha, param);
12207 PL_utf8_space = sv_dup_inc(proto_perl->Iutf8_space, param);
12208 PL_utf8_cntrl = sv_dup_inc(proto_perl->Iutf8_cntrl, param);
12209 PL_utf8_graph = sv_dup_inc(proto_perl->Iutf8_graph, param);
12210 PL_utf8_digit = sv_dup_inc(proto_perl->Iutf8_digit, param);
12211 PL_utf8_upper = sv_dup_inc(proto_perl->Iutf8_upper, param);
12212 PL_utf8_lower = sv_dup_inc(proto_perl->Iutf8_lower, param);
12213 PL_utf8_print = sv_dup_inc(proto_perl->Iutf8_print, param);
12214 PL_utf8_punct = sv_dup_inc(proto_perl->Iutf8_punct, param);
12215 PL_utf8_xdigit = sv_dup_inc(proto_perl->Iutf8_xdigit, param);
12216 PL_utf8_mark = sv_dup_inc(proto_perl->Iutf8_mark, param);
12217 PL_utf8_toupper = sv_dup_inc(proto_perl->Iutf8_toupper, param);
12218 PL_utf8_totitle = sv_dup_inc(proto_perl->Iutf8_totitle, param);
12219 PL_utf8_tolower = sv_dup_inc(proto_perl->Iutf8_tolower, param);
12220 PL_utf8_tofold = sv_dup_inc(proto_perl->Iutf8_tofold, param);
12221 PL_utf8_idstart = sv_dup_inc(proto_perl->Iutf8_idstart, param);
12222 PL_utf8_idcont = sv_dup_inc(proto_perl->Iutf8_idcont, param);
1d7c1841 12223
bd81e77b
NC
12224 /* Did the locale setup indicate UTF-8? */
12225 PL_utf8locale = proto_perl->Iutf8locale;
12226 /* Unicode features (see perlrun/-C) */
12227 PL_unicode = proto_perl->Iunicode;
1d7c1841 12228
bd81e77b
NC
12229 /* Pre-5.8 signals control */
12230 PL_signals = proto_perl->Isignals;
1d7c1841 12231
bd81e77b
NC
12232 /* times() ticks per second */
12233 PL_clocktick = proto_perl->Iclocktick;
1d7c1841 12234
bd81e77b
NC
12235 /* Recursion stopper for PerlIO_find_layer */
12236 PL_in_load_module = proto_perl->Iin_load_module;
8df990a8 12237
bd81e77b
NC
12238 /* sort() routine */
12239 PL_sort_RealCmp = proto_perl->Isort_RealCmp;
e5dd39fc 12240
bd81e77b
NC
12241 /* Not really needed/useful since the reenrant_retint is "volatile",
12242 * but do it for consistency's sake. */
12243 PL_reentrant_retint = proto_perl->Ireentrant_retint;
1d7c1841 12244
bd81e77b
NC
12245 /* Hooks to shared SVs and locks. */
12246 PL_sharehook = proto_perl->Isharehook;
12247 PL_lockhook = proto_perl->Ilockhook;
12248 PL_unlockhook = proto_perl->Iunlockhook;
12249 PL_threadhook = proto_perl->Ithreadhook;
eba16661 12250 PL_destroyhook = proto_perl->Idestroyhook;
1d7c1841 12251
bd81e77b
NC
12252#ifdef THREADS_HAVE_PIDS
12253 PL_ppid = proto_perl->Ippid;
12254#endif
1d7c1841 12255
bd81e77b 12256 /* swatch cache */
5c284bb0 12257 PL_last_swash_hv = NULL; /* reinits on demand */
bd81e77b
NC
12258 PL_last_swash_klen = 0;
12259 PL_last_swash_key[0]= '\0';
12260 PL_last_swash_tmps = (U8*)NULL;
12261 PL_last_swash_slen = 0;
1d7c1841 12262
bd81e77b
NC
12263 PL_glob_index = proto_perl->Iglob_index;
12264 PL_srand_called = proto_perl->Isrand_called;
05ec9bb3 12265
bd81e77b
NC
12266 if (proto_perl->Ipsig_pend) {
12267 Newxz(PL_psig_pend, SIG_SIZE, int);
12268 }
12269 else {
12270 PL_psig_pend = (int*)NULL;
12271 }
05ec9bb3 12272
d525a7b2
NC
12273 if (proto_perl->Ipsig_name) {
12274 Newx(PL_psig_name, 2 * SIG_SIZE, SV*);
12275 sv_dup_inc_multiple(proto_perl->Ipsig_name, PL_psig_name, 2 * SIG_SIZE,
538f2e76 12276 param);
d525a7b2 12277 PL_psig_ptr = PL_psig_name + SIG_SIZE;
bd81e77b
NC
12278 }
12279 else {
12280 PL_psig_ptr = (SV**)NULL;
12281 PL_psig_name = (SV**)NULL;
12282 }
05ec9bb3 12283
907b3e23 12284 /* intrpvar.h stuff */
1d7c1841 12285
bd81e77b
NC
12286 if (flags & CLONEf_COPY_STACKS) {
12287 /* next allocation will be PL_tmps_stack[PL_tmps_ix+1] */
907b3e23
DM
12288 PL_tmps_ix = proto_perl->Itmps_ix;
12289 PL_tmps_max = proto_perl->Itmps_max;
12290 PL_tmps_floor = proto_perl->Itmps_floor;
e92c6be8 12291 Newx(PL_tmps_stack, PL_tmps_max, SV*);
1d8a41fe
JD
12292 sv_dup_inc_multiple(proto_perl->Itmps_stack, PL_tmps_stack,
12293 PL_tmps_ix+1, param);
d2d73c3e 12294
bd81e77b 12295 /* next PUSHMARK() sets *(PL_markstack_ptr+1) */
907b3e23 12296 i = proto_perl->Imarkstack_max - proto_perl->Imarkstack;
bd81e77b 12297 Newxz(PL_markstack, i, I32);
907b3e23
DM
12298 PL_markstack_max = PL_markstack + (proto_perl->Imarkstack_max
12299 - proto_perl->Imarkstack);
12300 PL_markstack_ptr = PL_markstack + (proto_perl->Imarkstack_ptr
12301 - proto_perl->Imarkstack);
12302 Copy(proto_perl->Imarkstack, PL_markstack,
bd81e77b 12303 PL_markstack_ptr - PL_markstack + 1, I32);
d2d73c3e 12304
bd81e77b
NC
12305 /* next push_scope()/ENTER sets PL_scopestack[PL_scopestack_ix]
12306 * NOTE: unlike the others! */
907b3e23
DM
12307 PL_scopestack_ix = proto_perl->Iscopestack_ix;
12308 PL_scopestack_max = proto_perl->Iscopestack_max;
bd81e77b 12309 Newxz(PL_scopestack, PL_scopestack_max, I32);
907b3e23 12310 Copy(proto_perl->Iscopestack, PL_scopestack, PL_scopestack_ix, I32);
d419787a 12311
cbdd5331
JD
12312#ifdef DEBUGGING
12313 Newxz(PL_scopestack_name, PL_scopestack_max, const char *);
12314 Copy(proto_perl->Iscopestack_name, PL_scopestack_name, PL_scopestack_ix, const char *);
12315#endif
bd81e77b 12316 /* NOTE: si_dup() looks at PL_markstack */
907b3e23 12317 PL_curstackinfo = si_dup(proto_perl->Icurstackinfo, param);
d2d73c3e 12318
bd81e77b 12319 /* PL_curstack = PL_curstackinfo->si_stack; */
907b3e23
DM
12320 PL_curstack = av_dup(proto_perl->Icurstack, param);
12321 PL_mainstack = av_dup(proto_perl->Imainstack, param);
1d7c1841 12322
bd81e77b
NC
12323 /* next PUSHs() etc. set *(PL_stack_sp+1) */
12324 PL_stack_base = AvARRAY(PL_curstack);
907b3e23
DM
12325 PL_stack_sp = PL_stack_base + (proto_perl->Istack_sp
12326 - proto_perl->Istack_base);
bd81e77b 12327 PL_stack_max = PL_stack_base + AvMAX(PL_curstack);
1d7c1841 12328
bd81e77b
NC
12329 /* next SSPUSHFOO() sets PL_savestack[PL_savestack_ix]
12330 * NOTE: unlike the others! */
907b3e23
DM
12331 PL_savestack_ix = proto_perl->Isavestack_ix;
12332 PL_savestack_max = proto_perl->Isavestack_max;
bd81e77b
NC
12333 /*Newxz(PL_savestack, PL_savestack_max, ANY);*/
12334 PL_savestack = ss_dup(proto_perl, param);
12335 }
12336 else {
12337 init_stacks();
12338 ENTER; /* perl_destruct() wants to LEAVE; */
34394ecd
DM
12339
12340 /* although we're not duplicating the tmps stack, we should still
12341 * add entries for any SVs on the tmps stack that got cloned by a
12342 * non-refcount means (eg a temp in @_); otherwise they will be
12343 * orphaned
12344 */
907b3e23 12345 for (i = 0; i<= proto_perl->Itmps_ix; i++) {
daba3364
NC
12346 SV * const nsv = MUTABLE_SV(ptr_table_fetch(PL_ptr_table,
12347 proto_perl->Itmps_stack[i]));
34394ecd 12348 if (nsv && !SvREFCNT(nsv)) {
81041c50 12349 PUSH_EXTEND_MORTAL__SV_C(SvREFCNT_inc_simple(nsv));
34394ecd
DM
12350 }
12351 }
bd81e77b 12352 }
1d7c1841 12353
907b3e23 12354 PL_start_env = proto_perl->Istart_env; /* XXXXXX */
bd81e77b 12355 PL_top_env = &PL_start_env;
1d7c1841 12356
907b3e23 12357 PL_op = proto_perl->Iop;
4a4c6fe3 12358
a0714e2c 12359 PL_Sv = NULL;
bd81e77b 12360 PL_Xpv = (XPV*)NULL;
24792b8d 12361 my_perl->Ina = proto_perl->Ina;
1fcf4c12 12362
907b3e23
DM
12363 PL_statbuf = proto_perl->Istatbuf;
12364 PL_statcache = proto_perl->Istatcache;
12365 PL_statgv = gv_dup(proto_perl->Istatgv, param);
12366 PL_statname = sv_dup_inc(proto_perl->Istatname, param);
bd81e77b 12367#ifdef HAS_TIMES
907b3e23 12368 PL_timesbuf = proto_perl->Itimesbuf;
bd81e77b 12369#endif
1d7c1841 12370
907b3e23
DM
12371 PL_tainted = proto_perl->Itainted;
12372 PL_curpm = proto_perl->Icurpm; /* XXX No PMOP ref count */
12373 PL_rs = sv_dup_inc(proto_perl->Irs, param);
12374 PL_last_in_gv = gv_dup(proto_perl->Ilast_in_gv, param);
907b3e23
DM
12375 PL_defoutgv = gv_dup_inc(proto_perl->Idefoutgv, param);
12376 PL_chopset = proto_perl->Ichopset; /* XXX never deallocated */
12377 PL_toptarget = sv_dup_inc(proto_perl->Itoptarget, param);
12378 PL_bodytarget = sv_dup_inc(proto_perl->Ibodytarget, param);
12379 PL_formtarget = sv_dup(proto_perl->Iformtarget, param);
12380
12381 PL_restartop = proto_perl->Irestartop;
12382 PL_in_eval = proto_perl->Iin_eval;
12383 PL_delaymagic = proto_perl->Idelaymagic;
12384 PL_dirty = proto_perl->Idirty;
12385 PL_localizing = proto_perl->Ilocalizing;
12386
12387 PL_errors = sv_dup_inc(proto_perl->Ierrors, param);
4608196e 12388 PL_hv_fetch_ent_mh = NULL;
907b3e23 12389 PL_modcount = proto_perl->Imodcount;
5f66b61c 12390 PL_lastgotoprobe = NULL;
907b3e23 12391 PL_dumpindent = proto_perl->Idumpindent;
1d7c1841 12392
907b3e23
DM
12393 PL_sortcop = (OP*)any_dup(proto_perl->Isortcop, proto_perl);
12394 PL_sortstash = hv_dup(proto_perl->Isortstash, param);
12395 PL_firstgv = gv_dup(proto_perl->Ifirstgv, param);
12396 PL_secondgv = gv_dup(proto_perl->Isecondgv, param);
bd61b366 12397 PL_efloatbuf = NULL; /* reinits on demand */
bd81e77b 12398 PL_efloatsize = 0; /* reinits on demand */
d2d73c3e 12399
bd81e77b 12400 /* regex stuff */
1d7c1841 12401
bd81e77b
NC
12402 PL_screamfirst = NULL;
12403 PL_screamnext = NULL;
12404 PL_maxscream = -1; /* reinits on demand */
a0714e2c 12405 PL_lastscream = NULL;
1d7c1841 12406
1d7c1841 12407
907b3e23 12408 PL_regdummy = proto_perl->Iregdummy;
bd81e77b
NC
12409 PL_colorset = 0; /* reinits PL_colors[] */
12410 /*PL_colors[6] = {0,0,0,0,0,0};*/
1d7c1841 12411
84da74a7 12412
1d7c1841 12413
bd81e77b 12414 /* Pluggable optimizer */
907b3e23 12415 PL_peepp = proto_perl->Ipeepp;
f37b8c3f
VP
12416 /* op_free() hook */
12417 PL_opfreehook = proto_perl->Iopfreehook;
1d7c1841 12418
bd81e77b 12419 PL_stashcache = newHV();
1d7c1841 12420
b7185faf 12421 PL_watchaddr = (char **) ptr_table_fetch(PL_ptr_table,
907b3e23 12422 proto_perl->Iwatchaddr);
b7185faf
DM
12423 PL_watchok = PL_watchaddr ? * PL_watchaddr : NULL;
12424 if (PL_debug && PL_watchaddr) {
12425 PerlIO_printf(Perl_debug_log,
12426 "WATCHING: %"UVxf" cloned as %"UVxf" with value %"UVxf"\n",
907b3e23 12427 PTR2UV(proto_perl->Iwatchaddr), PTR2UV(PL_watchaddr),
b7185faf
DM
12428 PTR2UV(PL_watchok));
12429 }
12430
a3e6e81e
NC
12431 PL_registered_mros = hv_dup_inc(proto_perl->Iregistered_mros, param);
12432
bd81e77b
NC
12433 /* Call the ->CLONE method, if it exists, for each of the stashes
12434 identified by sv_dup() above.
12435 */
12436 while(av_len(param->stashes) != -1) {
85fbaab2 12437 HV* const stash = MUTABLE_HV(av_shift(param->stashes));
bd81e77b
NC
12438 GV* const cloner = gv_fetchmethod_autoload(stash, "CLONE", 0);
12439 if (cloner && GvCV(cloner)) {
12440 dSP;
12441 ENTER;
12442 SAVETMPS;
12443 PUSHMARK(SP);
6e449a3a 12444 mXPUSHs(newSVhek(HvNAME_HEK(stash)));
bd81e77b 12445 PUTBACK;
daba3364 12446 call_sv(MUTABLE_SV(GvCV(cloner)), G_DISCARD);
bd81e77b
NC
12447 FREETMPS;
12448 LEAVE;
12449 }
1d7c1841 12450 }
1d7c1841 12451
b0b93b3c
DM
12452 if (!(flags & CLONEf_KEEP_PTR_TABLE)) {
12453 ptr_table_free(PL_ptr_table);
12454 PL_ptr_table = NULL;
12455 }
12456
12457
bd81e77b 12458 SvREFCNT_dec(param->stashes);
1d7c1841 12459
bd81e77b
NC
12460 /* orphaned? eg threads->new inside BEGIN or use */
12461 if (PL_compcv && ! SvREFCNT(PL_compcv)) {
b37c2d43 12462 SvREFCNT_inc_simple_void(PL_compcv);
bd81e77b
NC
12463 SAVEFREESV(PL_compcv);
12464 }
dd2155a4 12465
bd81e77b
NC
12466 return my_perl;
12467}
1d7c1841 12468
bd81e77b 12469#endif /* USE_ITHREADS */
1d7c1841 12470
bd81e77b
NC
12471/*
12472=head1 Unicode Support
1d7c1841 12473
bd81e77b 12474=for apidoc sv_recode_to_utf8
1d7c1841 12475
bd81e77b
NC
12476The encoding is assumed to be an Encode object, on entry the PV
12477of the sv is assumed to be octets in that encoding, and the sv
12478will be converted into Unicode (and UTF-8).
1d7c1841 12479
bd81e77b
NC
12480If the sv already is UTF-8 (or if it is not POK), or if the encoding
12481is not a reference, nothing is done to the sv. If the encoding is not
12482an C<Encode::XS> Encoding object, bad things will happen.
12483(See F<lib/encoding.pm> and L<Encode>).
1d7c1841 12484
bd81e77b 12485The PV of the sv is returned.
1d7c1841 12486
bd81e77b 12487=cut */
1d7c1841 12488
bd81e77b
NC
12489char *
12490Perl_sv_recode_to_utf8(pTHX_ SV *sv, SV *encoding)
12491{
12492 dVAR;
7918f24d
NC
12493
12494 PERL_ARGS_ASSERT_SV_RECODE_TO_UTF8;
12495
bd81e77b
NC
12496 if (SvPOK(sv) && !SvUTF8(sv) && !IN_BYTES && SvROK(encoding)) {
12497 SV *uni;
12498 STRLEN len;
12499 const char *s;
12500 dSP;
12501 ENTER;
12502 SAVETMPS;
12503 save_re_context();
12504 PUSHMARK(sp);
12505 EXTEND(SP, 3);
12506 XPUSHs(encoding);
12507 XPUSHs(sv);
12508/*
12509 NI-S 2002/07/09
12510 Passing sv_yes is wrong - it needs to be or'ed set of constants
12511 for Encode::XS, while UTf-8 decode (currently) assumes a true value means
12512 remove converted chars from source.
1d7c1841 12513
bd81e77b 12514 Both will default the value - let them.
1d7c1841 12515
bd81e77b
NC
12516 XPUSHs(&PL_sv_yes);
12517*/
12518 PUTBACK;
12519 call_method("decode", G_SCALAR);
12520 SPAGAIN;
12521 uni = POPs;
12522 PUTBACK;
12523 s = SvPV_const(uni, len);
12524 if (s != SvPVX_const(sv)) {
12525 SvGROW(sv, len + 1);
12526 Move(s, SvPVX(sv), len + 1, char);
12527 SvCUR_set(sv, len);
12528 }
12529 FREETMPS;
12530 LEAVE;
12531 SvUTF8_on(sv);
12532 return SvPVX(sv);
389edf32 12533 }
bd81e77b
NC
12534 return SvPOKp(sv) ? SvPVX(sv) : NULL;
12535}
1d7c1841 12536
bd81e77b
NC
12537/*
12538=for apidoc sv_cat_decode
1d7c1841 12539
bd81e77b
NC
12540The encoding is assumed to be an Encode object, the PV of the ssv is
12541assumed to be octets in that encoding and decoding the input starts
12542from the position which (PV + *offset) pointed to. The dsv will be
12543concatenated the decoded UTF-8 string from ssv. Decoding will terminate
12544when the string tstr appears in decoding output or the input ends on
12545the PV of the ssv. The value which the offset points will be modified
12546to the last input position on the ssv.
1d7c1841 12547
bd81e77b 12548Returns TRUE if the terminator was found, else returns FALSE.
1d7c1841 12549
bd81e77b
NC
12550=cut */
12551
12552bool
12553Perl_sv_cat_decode(pTHX_ SV *dsv, SV *encoding,
12554 SV *ssv, int *offset, char *tstr, int tlen)
12555{
12556 dVAR;
12557 bool ret = FALSE;
7918f24d
NC
12558
12559 PERL_ARGS_ASSERT_SV_CAT_DECODE;
12560
bd81e77b
NC
12561 if (SvPOK(ssv) && SvPOK(dsv) && SvROK(encoding) && offset) {
12562 SV *offsv;
12563 dSP;
12564 ENTER;
12565 SAVETMPS;
12566 save_re_context();
12567 PUSHMARK(sp);
12568 EXTEND(SP, 6);
12569 XPUSHs(encoding);
12570 XPUSHs(dsv);
12571 XPUSHs(ssv);
6e449a3a
MHM
12572 offsv = newSViv(*offset);
12573 mXPUSHs(offsv);
12574 mXPUSHp(tstr, tlen);
bd81e77b
NC
12575 PUTBACK;
12576 call_method("cat_decode", G_SCALAR);
12577 SPAGAIN;
12578 ret = SvTRUE(TOPs);
12579 *offset = SvIV(offsv);
12580 PUTBACK;
12581 FREETMPS;
12582 LEAVE;
389edf32 12583 }
bd81e77b
NC
12584 else
12585 Perl_croak(aTHX_ "Invalid argument to sv_cat_decode");
12586 return ret;
1d7c1841 12587
bd81e77b 12588}
1d7c1841 12589
bd81e77b
NC
12590/* ---------------------------------------------------------------------
12591 *
12592 * support functions for report_uninit()
12593 */
1d7c1841 12594
bd81e77b
NC
12595/* the maxiumum size of array or hash where we will scan looking
12596 * for the undefined element that triggered the warning */
1d7c1841 12597
bd81e77b 12598#define FUV_MAX_SEARCH_SIZE 1000
1d7c1841 12599
bd81e77b
NC
12600/* Look for an entry in the hash whose value has the same SV as val;
12601 * If so, return a mortal copy of the key. */
1d7c1841 12602
bd81e77b 12603STATIC SV*
6c1b357c 12604S_find_hash_subscript(pTHX_ const HV *const hv, const SV *const val)
bd81e77b
NC
12605{
12606 dVAR;
12607 register HE **array;
12608 I32 i;
6c3182a5 12609
7918f24d
NC
12610 PERL_ARGS_ASSERT_FIND_HASH_SUBSCRIPT;
12611
bd81e77b
NC
12612 if (!hv || SvMAGICAL(hv) || !HvARRAY(hv) ||
12613 (HvTOTALKEYS(hv) > FUV_MAX_SEARCH_SIZE))
a0714e2c 12614 return NULL;
6c3182a5 12615
bd81e77b 12616 array = HvARRAY(hv);
6c3182a5 12617
bd81e77b
NC
12618 for (i=HvMAX(hv); i>0; i--) {
12619 register HE *entry;
12620 for (entry = array[i]; entry; entry = HeNEXT(entry)) {
12621 if (HeVAL(entry) != val)
12622 continue;
12623 if ( HeVAL(entry) == &PL_sv_undef ||
12624 HeVAL(entry) == &PL_sv_placeholder)
12625 continue;
12626 if (!HeKEY(entry))
a0714e2c 12627 return NULL;
bd81e77b
NC
12628 if (HeKLEN(entry) == HEf_SVKEY)
12629 return sv_mortalcopy(HeKEY_sv(entry));
a663657d 12630 return sv_2mortal(newSVhek(HeKEY_hek(entry)));
bd81e77b
NC
12631 }
12632 }
a0714e2c 12633 return NULL;
bd81e77b 12634}
6c3182a5 12635
bd81e77b
NC
12636/* Look for an entry in the array whose value has the same SV as val;
12637 * If so, return the index, otherwise return -1. */
6c3182a5 12638
bd81e77b 12639STATIC I32
6c1b357c 12640S_find_array_subscript(pTHX_ const AV *const av, const SV *const val)
bd81e77b 12641{
97aff369 12642 dVAR;
7918f24d
NC
12643
12644 PERL_ARGS_ASSERT_FIND_ARRAY_SUBSCRIPT;
12645
bd81e77b
NC
12646 if (!av || SvMAGICAL(av) || !AvARRAY(av) ||
12647 (AvFILLp(av) > FUV_MAX_SEARCH_SIZE))
12648 return -1;
57c6e6d2 12649
4a021917
AL
12650 if (val != &PL_sv_undef) {
12651 SV ** const svp = AvARRAY(av);
12652 I32 i;
12653
12654 for (i=AvFILLp(av); i>=0; i--)
12655 if (svp[i] == val)
12656 return i;
bd81e77b
NC
12657 }
12658 return -1;
12659}
15a5279a 12660
bd81e77b
NC
12661/* S_varname(): return the name of a variable, optionally with a subscript.
12662 * If gv is non-zero, use the name of that global, along with gvtype (one
12663 * of "$", "@", "%"); otherwise use the name of the lexical at pad offset
12664 * targ. Depending on the value of the subscript_type flag, return:
12665 */
bce260cd 12666
bd81e77b
NC
12667#define FUV_SUBSCRIPT_NONE 1 /* "@foo" */
12668#define FUV_SUBSCRIPT_ARRAY 2 /* "$foo[aindex]" */
12669#define FUV_SUBSCRIPT_HASH 3 /* "$foo{keyname}" */
12670#define FUV_SUBSCRIPT_WITHIN 4 /* "within @foo" */
bce260cd 12671
bd81e77b 12672STATIC SV*
6c1b357c
NC
12673S_varname(pTHX_ const GV *const gv, const char gvtype, PADOFFSET targ,
12674 const SV *const keyname, I32 aindex, int subscript_type)
bd81e77b 12675{
1d7c1841 12676
bd81e77b
NC
12677 SV * const name = sv_newmortal();
12678 if (gv) {
12679 char buffer[2];
12680 buffer[0] = gvtype;
12681 buffer[1] = 0;
1d7c1841 12682
bd81e77b 12683 /* as gv_fullname4(), but add literal '^' for $^FOO names */
66fe0623 12684
bd81e77b 12685 gv_fullname4(name, gv, buffer, 0);
1d7c1841 12686
bd81e77b
NC
12687 if ((unsigned int)SvPVX(name)[1] <= 26) {
12688 buffer[0] = '^';
12689 buffer[1] = SvPVX(name)[1] + 'A' - 1;
1d7c1841 12690
bd81e77b
NC
12691 /* Swap the 1 unprintable control character for the 2 byte pretty
12692 version - ie substr($name, 1, 1) = $buffer; */
12693 sv_insert(name, 1, 1, buffer, 2);
1d7c1841 12694 }
bd81e77b
NC
12695 }
12696 else {
289b91d9 12697 CV * const cv = find_runcv(NULL);
bd81e77b
NC
12698 SV *sv;
12699 AV *av;
1d7c1841 12700
bd81e77b 12701 if (!cv || !CvPADLIST(cv))
a0714e2c 12702 return NULL;
502c6561 12703 av = MUTABLE_AV((*av_fetch(CvPADLIST(cv), 0, FALSE)));
bd81e77b 12704 sv = *av_fetch(av, targ, FALSE);
f8503592 12705 sv_setpvn(name, SvPV_nolen_const(sv), SvCUR(sv));
bd81e77b 12706 }
1d7c1841 12707
bd81e77b 12708 if (subscript_type == FUV_SUBSCRIPT_HASH) {
561b68a9 12709 SV * const sv = newSV(0);
bd81e77b
NC
12710 *SvPVX(name) = '$';
12711 Perl_sv_catpvf(aTHX_ name, "{%s}",
12712 pv_display(sv,SvPVX_const(keyname), SvCUR(keyname), 0, 32));
12713 SvREFCNT_dec(sv);
12714 }
12715 else if (subscript_type == FUV_SUBSCRIPT_ARRAY) {
12716 *SvPVX(name) = '$';
12717 Perl_sv_catpvf(aTHX_ name, "[%"IVdf"]", (IV)aindex);
12718 }
84335ee9
NC
12719 else if (subscript_type == FUV_SUBSCRIPT_WITHIN) {
12720 /* We know that name has no magic, so can use 0 instead of SV_GMAGIC */
12721 Perl_sv_insert_flags(aTHX_ name, 0, 0, STR_WITH_LEN("within "), 0);
12722 }
1d7c1841 12723
bd81e77b
NC
12724 return name;
12725}
1d7c1841 12726
1d7c1841 12727
bd81e77b
NC
12728/*
12729=for apidoc find_uninit_var
1d7c1841 12730
bd81e77b
NC
12731Find the name of the undefined variable (if any) that caused the operator o
12732to issue a "Use of uninitialized value" warning.
12733If match is true, only return a name if it's value matches uninit_sv.
12734So roughly speaking, if a unary operator (such as OP_COS) generates a
12735warning, then following the direct child of the op may yield an
12736OP_PADSV or OP_GV that gives the name of the undefined variable. On the
12737other hand, with OP_ADD there are two branches to follow, so we only print
12738the variable name if we get an exact match.
1d7c1841 12739
bd81e77b 12740The name is returned as a mortal SV.
1d7c1841 12741
bd81e77b
NC
12742Assumes that PL_op is the op that originally triggered the error, and that
12743PL_comppad/PL_curpad points to the currently executing pad.
1d7c1841 12744
bd81e77b
NC
12745=cut
12746*/
1d7c1841 12747
bd81e77b 12748STATIC SV *
6c1b357c
NC
12749S_find_uninit_var(pTHX_ const OP *const obase, const SV *const uninit_sv,
12750 bool match)
bd81e77b
NC
12751{
12752 dVAR;
12753 SV *sv;
6c1b357c
NC
12754 const GV *gv;
12755 const OP *o, *o2, *kid;
1d7c1841 12756
bd81e77b
NC
12757 if (!obase || (match && (!uninit_sv || uninit_sv == &PL_sv_undef ||
12758 uninit_sv == &PL_sv_placeholder)))
a0714e2c 12759 return NULL;
1d7c1841 12760
bd81e77b 12761 switch (obase->op_type) {
1d7c1841 12762
bd81e77b
NC
12763 case OP_RV2AV:
12764 case OP_RV2HV:
12765 case OP_PADAV:
12766 case OP_PADHV:
12767 {
12768 const bool pad = (obase->op_type == OP_PADAV || obase->op_type == OP_PADHV);
12769 const bool hash = (obase->op_type == OP_PADHV || obase->op_type == OP_RV2HV);
12770 I32 index = 0;
a0714e2c 12771 SV *keysv = NULL;
bd81e77b 12772 int subscript_type = FUV_SUBSCRIPT_WITHIN;
1d7c1841 12773
bd81e77b
NC
12774 if (pad) { /* @lex, %lex */
12775 sv = PAD_SVl(obase->op_targ);
a0714e2c 12776 gv = NULL;
bd81e77b
NC
12777 }
12778 else {
12779 if (cUNOPx(obase)->op_first->op_type == OP_GV) {
12780 /* @global, %global */
12781 gv = cGVOPx_gv(cUNOPx(obase)->op_first);
12782 if (!gv)
12783 break;
daba3364 12784 sv = hash ? MUTABLE_SV(GvHV(gv)): MUTABLE_SV(GvAV(gv));
bd81e77b
NC
12785 }
12786 else /* @{expr}, %{expr} */
12787 return find_uninit_var(cUNOPx(obase)->op_first,
12788 uninit_sv, match);
12789 }
1d7c1841 12790
bd81e77b
NC
12791 /* attempt to find a match within the aggregate */
12792 if (hash) {
85fbaab2 12793 keysv = find_hash_subscript((const HV*)sv, uninit_sv);
bd81e77b
NC
12794 if (keysv)
12795 subscript_type = FUV_SUBSCRIPT_HASH;
12796 }
12797 else {
502c6561 12798 index = find_array_subscript((const AV *)sv, uninit_sv);
bd81e77b
NC
12799 if (index >= 0)
12800 subscript_type = FUV_SUBSCRIPT_ARRAY;
12801 }
1d7c1841 12802
bd81e77b
NC
12803 if (match && subscript_type == FUV_SUBSCRIPT_WITHIN)
12804 break;
1d7c1841 12805
bd81e77b
NC
12806 return varname(gv, hash ? '%' : '@', obase->op_targ,
12807 keysv, index, subscript_type);
12808 }
1d7c1841 12809
bd81e77b
NC
12810 case OP_PADSV:
12811 if (match && PAD_SVl(obase->op_targ) != uninit_sv)
12812 break;
a0714e2c
SS
12813 return varname(NULL, '$', obase->op_targ,
12814 NULL, 0, FUV_SUBSCRIPT_NONE);
1d7c1841 12815
bd81e77b
NC
12816 case OP_GVSV:
12817 gv = cGVOPx_gv(obase);
12818 if (!gv || (match && GvSV(gv) != uninit_sv))
12819 break;
a0714e2c 12820 return varname(gv, '$', 0, NULL, 0, FUV_SUBSCRIPT_NONE);
1d7c1841 12821
bd81e77b
NC
12822 case OP_AELEMFAST:
12823 if (obase->op_flags & OPf_SPECIAL) { /* lexical array */
12824 if (match) {
12825 SV **svp;
502c6561 12826 AV *av = MUTABLE_AV(PAD_SV(obase->op_targ));
bd81e77b
NC
12827 if (!av || SvRMAGICAL(av))
12828 break;
12829 svp = av_fetch(av, (I32)obase->op_private, FALSE);
12830 if (!svp || *svp != uninit_sv)
12831 break;
12832 }
a0714e2c
SS
12833 return varname(NULL, '$', obase->op_targ,
12834 NULL, (I32)obase->op_private, FUV_SUBSCRIPT_ARRAY);
bd81e77b
NC
12835 }
12836 else {
12837 gv = cGVOPx_gv(obase);
12838 if (!gv)
12839 break;
12840 if (match) {
12841 SV **svp;
6c1b357c 12842 AV *const av = GvAV(gv);
bd81e77b
NC
12843 if (!av || SvRMAGICAL(av))
12844 break;
12845 svp = av_fetch(av, (I32)obase->op_private, FALSE);
12846 if (!svp || *svp != uninit_sv)
12847 break;
12848 }
12849 return varname(gv, '$', 0,
a0714e2c 12850 NULL, (I32)obase->op_private, FUV_SUBSCRIPT_ARRAY);
bd81e77b
NC
12851 }
12852 break;
1d7c1841 12853
bd81e77b
NC
12854 case OP_EXISTS:
12855 o = cUNOPx(obase)->op_first;
12856 if (!o || o->op_type != OP_NULL ||
12857 ! (o->op_targ == OP_AELEM || o->op_targ == OP_HELEM))
12858 break;
12859 return find_uninit_var(cBINOPo->op_last, uninit_sv, match);
a2efc822 12860
bd81e77b
NC
12861 case OP_AELEM:
12862 case OP_HELEM:
12863 if (PL_op == obase)
12864 /* $a[uninit_expr] or $h{uninit_expr} */
12865 return find_uninit_var(cBINOPx(obase)->op_last, uninit_sv, match);
081fc587 12866
a0714e2c 12867 gv = NULL;
bd81e77b
NC
12868 o = cBINOPx(obase)->op_first;
12869 kid = cBINOPx(obase)->op_last;
8cf8f3d1 12870
bd81e77b 12871 /* get the av or hv, and optionally the gv */
a0714e2c 12872 sv = NULL;
bd81e77b
NC
12873 if (o->op_type == OP_PADAV || o->op_type == OP_PADHV) {
12874 sv = PAD_SV(o->op_targ);
12875 }
12876 else if ((o->op_type == OP_RV2AV || o->op_type == OP_RV2HV)
12877 && cUNOPo->op_first->op_type == OP_GV)
12878 {
12879 gv = cGVOPx_gv(cUNOPo->op_first);
12880 if (!gv)
12881 break;
daba3364
NC
12882 sv = o->op_type
12883 == OP_RV2HV ? MUTABLE_SV(GvHV(gv)) : MUTABLE_SV(GvAV(gv));
bd81e77b
NC
12884 }
12885 if (!sv)
12886 break;
12887
12888 if (kid && kid->op_type == OP_CONST && SvOK(cSVOPx_sv(kid))) {
12889 /* index is constant */
12890 if (match) {
12891 if (SvMAGICAL(sv))
12892 break;
12893 if (obase->op_type == OP_HELEM) {
85fbaab2 12894 HE* he = hv_fetch_ent(MUTABLE_HV(sv), cSVOPx_sv(kid), 0, 0);
bd81e77b
NC
12895 if (!he || HeVAL(he) != uninit_sv)
12896 break;
12897 }
12898 else {
502c6561 12899 SV * const * const svp = av_fetch(MUTABLE_AV(sv), SvIV(cSVOPx_sv(kid)), FALSE);
bd81e77b
NC
12900 if (!svp || *svp != uninit_sv)
12901 break;
12902 }
12903 }
12904 if (obase->op_type == OP_HELEM)
12905 return varname(gv, '%', o->op_targ,
12906 cSVOPx_sv(kid), 0, FUV_SUBSCRIPT_HASH);
12907 else
a0714e2c 12908 return varname(gv, '@', o->op_targ, NULL,
bd81e77b 12909 SvIV(cSVOPx_sv(kid)), FUV_SUBSCRIPT_ARRAY);
bd81e77b
NC
12910 }
12911 else {
12912 /* index is an expression;
12913 * attempt to find a match within the aggregate */
12914 if (obase->op_type == OP_HELEM) {
85fbaab2 12915 SV * const keysv = find_hash_subscript((const HV*)sv, uninit_sv);
bd81e77b
NC
12916 if (keysv)
12917 return varname(gv, '%', o->op_targ,
12918 keysv, 0, FUV_SUBSCRIPT_HASH);
12919 }
12920 else {
502c6561
NC
12921 const I32 index
12922 = find_array_subscript((const AV *)sv, uninit_sv);
bd81e77b
NC
12923 if (index >= 0)
12924 return varname(gv, '@', o->op_targ,
a0714e2c 12925 NULL, index, FUV_SUBSCRIPT_ARRAY);
bd81e77b
NC
12926 }
12927 if (match)
12928 break;
12929 return varname(gv,
12930 (o->op_type == OP_PADAV || o->op_type == OP_RV2AV)
12931 ? '@' : '%',
a0714e2c 12932 o->op_targ, NULL, 0, FUV_SUBSCRIPT_WITHIN);
f284b03f 12933 }
bd81e77b 12934 break;
dc507217 12935
bd81e77b
NC
12936 case OP_AASSIGN:
12937 /* only examine RHS */
12938 return find_uninit_var(cBINOPx(obase)->op_first, uninit_sv, match);
6d26897e 12939
bd81e77b
NC
12940 case OP_OPEN:
12941 o = cUNOPx(obase)->op_first;
12942 if (o->op_type == OP_PUSHMARK)
12943 o = o->op_sibling;
1d7c1841 12944
bd81e77b
NC
12945 if (!o->op_sibling) {
12946 /* one-arg version of open is highly magical */
a0ae6670 12947
bd81e77b
NC
12948 if (o->op_type == OP_GV) { /* open FOO; */
12949 gv = cGVOPx_gv(o);
12950 if (match && GvSV(gv) != uninit_sv)
12951 break;
12952 return varname(gv, '$', 0,
a0714e2c 12953 NULL, 0, FUV_SUBSCRIPT_NONE);
bd81e77b
NC
12954 }
12955 /* other possibilities not handled are:
12956 * open $x; or open my $x; should return '${*$x}'
12957 * open expr; should return '$'.expr ideally
12958 */
12959 break;
12960 }
12961 goto do_op;
ccfc67b7 12962
bd81e77b
NC
12963 /* ops where $_ may be an implicit arg */
12964 case OP_TRANS:
12965 case OP_SUBST:
12966 case OP_MATCH:
12967 if ( !(obase->op_flags & OPf_STACKED)) {
12968 if (uninit_sv == ((obase->op_private & OPpTARGET_MY)
12969 ? PAD_SVl(obase->op_targ)
12970 : DEFSV))
12971 {
12972 sv = sv_newmortal();
76f68e9b 12973 sv_setpvs(sv, "$_");
bd81e77b
NC
12974 return sv;
12975 }
12976 }
12977 goto do_op;
9f4817db 12978
bd81e77b
NC
12979 case OP_PRTF:
12980 case OP_PRINT:
3ef1310e 12981 case OP_SAY:
fa8d1836 12982 match = 1; /* print etc can return undef on defined args */
bd81e77b
NC
12983 /* skip filehandle as it can't produce 'undef' warning */
12984 o = cUNOPx(obase)->op_first;
12985 if ((obase->op_flags & OPf_STACKED) && o->op_type == OP_PUSHMARK)
12986 o = o->op_sibling->op_sibling;
12987 goto do_op2;
9f4817db 12988
9f4817db 12989
50edf520 12990 case OP_ENTEREVAL: /* could be eval $undef or $x='$undef'; eval $x */
bd81e77b 12991 case OP_RV2SV:
8b0dea50
DM
12992 case OP_CUSTOM: /* XS or custom code could trigger random warnings */
12993
12994 /* the following ops are capable of returning PL_sv_undef even for
12995 * defined arg(s) */
12996
12997 case OP_BACKTICK:
12998 case OP_PIPE_OP:
12999 case OP_FILENO:
13000 case OP_BINMODE:
13001 case OP_TIED:
13002 case OP_GETC:
13003 case OP_SYSREAD:
13004 case OP_SEND:
13005 case OP_IOCTL:
13006 case OP_SOCKET:
13007 case OP_SOCKPAIR:
13008 case OP_BIND:
13009 case OP_CONNECT:
13010 case OP_LISTEN:
13011 case OP_ACCEPT:
13012 case OP_SHUTDOWN:
13013 case OP_SSOCKOPT:
13014 case OP_GETPEERNAME:
13015 case OP_FTRREAD:
13016 case OP_FTRWRITE:
13017 case OP_FTREXEC:
13018 case OP_FTROWNED:
13019 case OP_FTEREAD:
13020 case OP_FTEWRITE:
13021 case OP_FTEEXEC:
13022 case OP_FTEOWNED:
13023 case OP_FTIS:
13024 case OP_FTZERO:
13025 case OP_FTSIZE:
13026 case OP_FTFILE:
13027 case OP_FTDIR:
13028 case OP_FTLINK:
13029 case OP_FTPIPE:
13030 case OP_FTSOCK:
13031 case OP_FTBLK:
13032 case OP_FTCHR:
13033 case OP_FTTTY:
13034 case OP_FTSUID:
13035 case OP_FTSGID:
13036 case OP_FTSVTX:
13037 case OP_FTTEXT:
13038 case OP_FTBINARY:
13039 case OP_FTMTIME:
13040 case OP_FTATIME:
13041 case OP_FTCTIME:
13042 case OP_READLINK:
13043 case OP_OPEN_DIR:
13044 case OP_READDIR:
13045 case OP_TELLDIR:
13046 case OP_SEEKDIR:
13047 case OP_REWINDDIR:
13048 case OP_CLOSEDIR:
13049 case OP_GMTIME:
13050 case OP_ALARM:
13051 case OP_SEMGET:
13052 case OP_GETLOGIN:
13053 case OP_UNDEF:
13054 case OP_SUBSTR:
13055 case OP_AEACH:
13056 case OP_EACH:
13057 case OP_SORT:
13058 case OP_CALLER:
13059 case OP_DOFILE:
fa8d1836
DM
13060 case OP_PROTOTYPE:
13061 case OP_NCMP:
13062 case OP_SMARTMATCH:
13063 case OP_UNPACK:
13064 case OP_SYSOPEN:
13065 case OP_SYSSEEK:
8b0dea50 13066 match = 1;
bd81e77b 13067 goto do_op;
9f4817db 13068
7697b7e7
DM
13069 case OP_ENTERSUB:
13070 case OP_GOTO:
a2fb3d36
DM
13071 /* XXX tmp hack: these two may call an XS sub, and currently
13072 XS subs don't have a SUB entry on the context stack, so CV and
13073 pad determination goes wrong, and BAD things happen. So, just
13074 don't try to determine the value under those circumstances.
7697b7e7
DM
13075 Need a better fix at dome point. DAPM 11/2007 */
13076 break;
13077
4f187fc9
VP
13078 case OP_FLIP:
13079 case OP_FLOP:
13080 {
13081 GV * const gv = gv_fetchpvs(".", GV_NOTQUAL, SVt_PV);
13082 if (gv && GvSV(gv) == uninit_sv)
13083 return newSVpvs_flags("$.", SVs_TEMP);
13084 goto do_op;
13085 }
8b0dea50 13086
cc4b8646
DM
13087 case OP_POS:
13088 /* def-ness of rval pos() is independent of the def-ness of its arg */
13089 if ( !(obase->op_flags & OPf_MOD))
13090 break;
13091
bd81e77b
NC
13092 case OP_SCHOMP:
13093 case OP_CHOMP:
13094 if (SvROK(PL_rs) && uninit_sv == SvRV(PL_rs))
84bafc02 13095 return newSVpvs_flags("${$/}", SVs_TEMP);
5f66b61c 13096 /*FALLTHROUGH*/
5d170f3a 13097
bd81e77b
NC
13098 default:
13099 do_op:
13100 if (!(obase->op_flags & OPf_KIDS))
13101 break;
13102 o = cUNOPx(obase)->op_first;
13103
13104 do_op2:
13105 if (!o)
13106 break;
f9893866 13107
bd81e77b
NC
13108 /* if all except one arg are constant, or have no side-effects,
13109 * or are optimized away, then it's unambiguous */
5f66b61c 13110 o2 = NULL;
bd81e77b 13111 for (kid=o; kid; kid = kid->op_sibling) {
e15d5972
AL
13112 if (kid) {
13113 const OPCODE type = kid->op_type;
13114 if ( (type == OP_CONST && SvOK(cSVOPx_sv(kid)))
13115 || (type == OP_NULL && ! (kid->op_flags & OPf_KIDS))
13116 || (type == OP_PUSHMARK)
bd81e77b 13117 )
bd81e77b 13118 continue;
e15d5972 13119 }
bd81e77b 13120 if (o2) { /* more than one found */
5f66b61c 13121 o2 = NULL;
bd81e77b
NC
13122 break;
13123 }
13124 o2 = kid;
13125 }
13126 if (o2)
13127 return find_uninit_var(o2, uninit_sv, match);
7a5fa8a2 13128
bd81e77b
NC
13129 /* scan all args */
13130 while (o) {
13131 sv = find_uninit_var(o, uninit_sv, 1);
13132 if (sv)
13133 return sv;
13134 o = o->op_sibling;
d0063567 13135 }
bd81e77b 13136 break;
f9893866 13137 }
a0714e2c 13138 return NULL;
9f4817db
JH
13139}
13140
220e2d4e 13141
bd81e77b
NC
13142/*
13143=for apidoc report_uninit
68795e93 13144
bd81e77b 13145Print appropriate "Use of uninitialized variable" warning
220e2d4e 13146
bd81e77b
NC
13147=cut
13148*/
220e2d4e 13149
bd81e77b 13150void
b3dbd76e 13151Perl_report_uninit(pTHX_ const SV *uninit_sv)
220e2d4e 13152{
97aff369 13153 dVAR;
bd81e77b 13154 if (PL_op) {
a0714e2c 13155 SV* varname = NULL;
bd81e77b
NC
13156 if (uninit_sv) {
13157 varname = find_uninit_var(PL_op, uninit_sv,0);
13158 if (varname)
13159 sv_insert(varname, 0, 0, " ", 1);
13160 }
13161 Perl_warner(aTHX_ packWARN(WARN_UNINITIALIZED), PL_warn_uninit,
13162 varname ? SvPV_nolen_const(varname) : "",
13163 " in ", OP_DESC(PL_op));
220e2d4e 13164 }
a73e8557 13165 else
bd81e77b
NC
13166 Perl_warner(aTHX_ packWARN(WARN_UNINITIALIZED), PL_warn_uninit,
13167 "", "", "");
220e2d4e 13168}
f9893866 13169
241d1a3b
NC
13170/*
13171 * Local variables:
13172 * c-indentation-style: bsd
13173 * c-basic-offset: 4
13174 * indent-tabs-mode: t
13175 * End:
13176 *
37442d52
RGS
13177 * ex: set ts=8 sts=4 sw=4 noet:
13178 */